International Journal of Medical Science and Public Health Research
63
https://ijmsphr.com/index.php/ijmsphr
TYPE
Original Research
PAGE NO.
63-72
DOI
10.37547/ijmsphr/Volume06Issue05-04
OPEN ACCESS
SUBMITED
19 March 2025
ACCEPTED
14 April 2025
PUBLISHED
11 May 2025
VOLUME
Vol.06 Issue 05 2025
CITATION
Mazharul Islam Tusher, Md Refat Hossain, Arjina Akter, Md Rayhan
Hassan Mahin, Sharmin Sultana Akhi, MD Sajedul Karim Chy, Mahfuz
Haider, Sadia Akter, Md Minzamul Hasan, & Mujiba Shaima. (2025).
Deep Learning Meets Early Diagnosis: A Hybrid CNN-DNN Framework
for Lung Cancer Prediction and Clinical Translation. International
Journal of Medical Science and Public Health Research, 6(05), 63
–
72.
https://doi.org/10.37547/ijmsphr/Volume06Issue05-04
COPYRIGHT
© 2025 Original content from this work may be used under the terms
of the creative commons attributes 4.0 License.
Deep Learning Meets Early
Diagnosis: A Hybrid CNN-
DNN Framework for Lung
Cancer Prediction and
Clinical Translation
Mazharul Islam Tusher
Department of Computer Science, Monroe College, New Rochelle, New
York, USA
Md Refat Hossain
Master of Business Administration, Westcliff University, USA
Arjina Akter
Department of Public Health, Central Michigan University, Mount
Pleasant, Michigan, USA
Md Rayhan Hassan Mahin
Department of Computer Science, Monroe University, New Rochelle,
USA
Sharmin Sultana Akhi
Department of Computer Science, Monroe University, USA
MD Sajedul Karim Chy
Department of Business Administration, Washington University of
Science and Technology, USA
Mahfuz Haider
Clinical Operations Analyst, Department of Clinical Operations, University
of Virginia Physicians Group
Sadia Akter
Department of Business Administration, International American
University, USA
Md Minzamul Hasan
Doctor of Business Administration (DBA), College of Business, Westcliff
University, USA
Mujiba Shaima
Department of Computer Science. Monroe University, NY
Abstract:
Early detection of lung cancer significantly improves
patient survival yet remains a challenge due to the
International Journal of Medical Science and Public Health Research
64
https://ijmsphr.com/index.php/ijmsphr
subtle nature of early-stage radiological features. This
study proposes a multimodal deep learning framework
that combines convolutional neural networks (CNN)
with dense neural networks (DNN) to enhance early-
stage lung cancer prediction. A curated dataset
comprising 2,000 patient records with CT scans and
clinical metadata was preprocessed and used to train
multiple models. The hybrid CNN-DNN model achieved
the highest performance with an accuracy of 96.4%,
precision of 95.8%, recall of 97.1%, F1-score of 96.4%,
and AUC of 0.982, outperforming both traditional
machine learning models and standalone CNNs. The
integration of imaging and clinical features led to
robust and accurate predictions, demonstrating strong
potential for real-world clinical application. The results
support the deployment of such AI-driven tools in
diagnostic workflows to facilitate timely and accurate
lung cancer detection.
Keywords:
Lung cancer detection, deep learning,
convolutional neural networks (CNN), dense neural
networks (DNN), early diagnosis, medical imaging,
clinical integration, artificial intelligence (AI), hybrid
model, computer-aided diagnosis.
Introduction:
Lung cancer remains one of the most
prevalent and deadly forms of cancer worldwide,
accounting for an estimated 1.8 million deaths
annually (World Health Organization [WHO], 2021).
Despite advancements in medical imaging and clinical
diagnostics, the survival rate of lung cancer patients
remains critically low, primarily due to delayed
detection. Early-stage lung cancer, if diagnosed and
treated promptly, offers a significantly higher survival
probability compared to late-stage cases. However,
current diagnostic procedures, such as CT imaging and
biopsy, often rely heavily on radiologist interpretation
and invasive measures, which can be time-consuming,
expensive, and prone to inter-observer variability.
In recent years, artificial intelligence (AI), particularly
deep learning (DL), has emerged as a transformative
tool in medical imaging and diagnostics. Deep learning
algorithms, especially convolutional neural networks
(CNNs), have demonstrated remarkable performance
in visual recognition tasks and are now being applied
to disease prediction and classification using
radiographic data. The integration of deep learning
into lung cancer detection presents a promising
approach to automate and enhance the accuracy of
early diagnosis, potentially improving patient outcomes
while reducing healthcare costs.
This study aims to develop a comprehensive deep
learning-based framework for early-stage lung cancer
prediction using a hybrid model that incorporates both
imaging data (e.g., CT scans) and structured clinical
information. By leveraging state-of-the-art CNN
architectures and data fusion techniques, the research
seeks to identify at-risk individuals at an early stage,
thereby contributing to better clinical decision-making
and improved survival rates.
LITERATURE REVIEW
The application of machine learning (ML) and deep
learning
in
oncology
has
seen
considerable
advancement over the past decade. Numerous studies
have explored the utility of DL models in automating
lung cancer detection through the analysis of CT imaging
data. For instance, Shen et al. (2017) proposed a multi-
crop CNN model to detect lung nodules and achieved
promising results, highlighting the potential of deep
neural networks in radiology. Similarly, Ardila et al.
(2019) developed an end-to-end DL system capable of
predicting lung cancer risk from raw CT scans with
performance comparable to that of experienced
radiologists.
Several studies have utilized transfer learning
approaches using pretrained models such as VGG16,
ResNet, and InceptionNet to improve classification
accuracy. For example, Wang et al. (2020)
demonstrated that fine-tuning ResNet50 significantly
enhanced diagnostic accuracy for lung nodule
classification, while Choi et al. (2021) integrated clinical
features with CNN outputs to create a more holistic
model. These findings suggest that incorporating both
image and non-image data can substantially improve
predictive performance.
The integration of clinical data such as age, gender,
smoking history, and family history into deep learning
pipelines is gaining traction as a means to contextualize
image-based predictions (Esteva et al., 2019). This
hybrid approach aligns with the principles of precision
medicine, allowing for personalized risk assessments
based on a combination of phenotypic and demographic
International Journal of Medical Science and Public Health Research
65
https://ijmsphr.com/index.php/ijmsphr
attributes.
Despite these advancements, many existing studies
focus primarily on binary classification or nodule
detection without emphasizing early-stage diagnosis.
Moreover, there is a lack of robust comparative
analysis among different deep learning architectures
using real-world, multimodal datasets. This research
addresses these gaps by developing and evaluating a
hybrid CNN-DNN model that fuses imaging and clinical
data for early-stage lung cancer prediction, offering
insights into clinical applicability and model
explainability.
METHODOLOGY
This study aims to develop a robust deep learning
framework to predict early-stage lung cancer using
both clinical data and medical imaging. The
methodology consists of six major phases: data
collection, data preprocessing, feature selection,
feature engineering, model development, and model
evaluation. Each phase is meticulously designed to
ensure accuracy, reproducibility, and applicability of
the predictive model in real-world clinical scenarios.
Data Collection
The initial phase of this research involves the
acquisition of high-quality datasets that can effectively
represent characteristics associated with early-stage
lung cancer. Multiple sources of data were combined to
ensure a diverse, rich, and balanced representation.
This includes publicly available medical image
repositories and structured clinical data. The primary
dataset used in this study is the LIDC-IDRI (Lung Image
Database Consortium and Image Database Resource
Initiative), which is among the most widely used
resources in lung cancer research. It contains over 1,000
thoracic computed tomography (CT) scans with
annotations
from
four
experienced
thoracic
radiologists. These annotations describe the size, shape,
texture, and likelihood of malignancy of lung nodules,
which are crucial in the detection of early-stage tumors.
In addition to image-based data, structured clinical
datasets sourced from Kaggle were utilized. These
datasets consist of patient demographic and clinical
history, such as age, gender, smoking habits, anxiety
levels, presence of chronic diseases, and family history
of lung cancer.
To enrich and diversify the sample space, additional
anonymized patient records were considered from
regional hospitals, subject to ethical approval and data
sharing agreements. These clinical records contain
essential diagnostic information that supplements
image data and enhances the learning capacity of hybrid
deep learning models.
Below table 1 is a summary of the datasets used in this research:
Dataset
Name
Type
Source
Data
Format
No. of
Samples
Key Attributes
LIDC-IDRI
CT Scans
The Cancer
Imaging Archive
(TCIA)
DICOM
1,018
Nodule annotations,
malignancy rating, lung
segmentation
Lung Cancer
Dataset
Clinical
Tabular
Kaggle
CSV
1,000
Age, gender, smoking status,
anxiety, chronic disease,
diagnosis
Hospital
Patient
Records
Clinical
Tabular
Regional Medical
Institutions
CSV/Excel
600
Blood test results, oxygen
levels, diagnostic comments,
family history
International Journal of Medical Science and Public Health Research
66
https://ijmsphr.com/index.php/ijmsphr
These datasets collectively offer a comprehensive
foundation to develop a model capable of detecting
lung cancer at its earliest stages.
Data Preprocessing
Following data acquisition, preprocessing plays a
critical role in transforming raw, unstructured, and
inconsistent data into a standardized format suitable
for deep learning algorithms. Clinical datasets often
contain missing entries due to incomplete hospital
records or survey responses. These missing values
were imputed using statistical methods such as mean
imputation for numerical features and mode
substitution for categorical attributes. Outliers were
detected using boxplots and z-scores and were treated
or removed depending on their relevance to lung
cancer symptoms.
The categorical variables, including gender and
smoking history, were converted into numerical
representations through encoding techniques. Label
encoding was applied to binary variables while one-hot
encoding was employed for variables with more than
two categories. Continuous features such as age and
blood pressure were normalized using min-max scaling
to ensure that the model assigns equal weight across
features.
In the case of image data, several preprocessing steps
were applied. The raw CT images from the LIDC-IDRI
dataset were in DICOM format and had varying
resolutions. Each image was resized to a consistent
resolution of 224x224 pixels to maintain uniformity for
model input. Pixel intensity normalization was
conducted to scale values between 0 and 1. Denoising
techniques such as Gaussian filtering were used to
reduce visual noise that might mislead the model. Lung
regions were segmented using thresholding followed
by morphological operations to extract relevant
nodular structures and suppress irrelevant regions like
bones or background air. Furthermore, data
augmentation strategies including rotation, flipping,
cropping, and zooming were employed to synthetically
enlarge the training dataset and prevent overfitting.
Feature Selection
In order to improve model interpretability and
computational efficiency, feature selection was carried
out to identify the most significant attributes
contributing to early-stage lung cancer prediction. For
clinical data, univariate statistical tests such as Chi-
square and ANOVA were performed to assess the
correlation between individual features and the target
label. These tests helped in eliminating features that
were either redundant or had weak predictive power.
Recursive Feature Elimination (RFE), a wrapper-based
method, was then applied using a base logistic
regression and random forest classifier to further
narrow down the most relevant features. The features
retained after these steps included patient age, years of
smoking, chronic disease presence, anxiety level, and
family history of cancer.
For image data, deep convolutional layers inherently
learn discriminative features during training, hence
manual selection was not required. Instead,
intermediate feature maps from CNN layers were
visualized to understand which regions the model
focused on, thereby confirming the relevance of
features learned.
Feature Engineering
To enhance model performance and represent complex
relationships within the data, feature engineering
techniques were employed. In the clinical data,
interaction terms were created by combining features
such as smoking status and age to capture compound
effects that could be more indicative of cancer risk than
individual features alone. Polynomial transformations
were used on continuous variables to expose non-linear
trends in the data.
Advanced radiomic features were extracted from the
segmented CT scans using texture analysis, edge
detection, and histogram of oriented gradients (HOG).
These features included skewness, kurtosis, entropy,
and energy, which provided insights into the shape and
texture of lung nodules. To manage high dimensionality
arising from engineered features, dimensionality
reduction techniques such as Principal Component
Analysis (PCA) and t-distributed Stochastic Neighbor
Embedding (t-SNE) were used for feature space
optimization.
The processed features from both clinical and image
domains were standardized and combined in cases
where multi-modal deep learning architecture was
applied, ensuring feature scale consistency and avoiding
model bias toward one data type
International Journal of Medical Science and Public Health Research
67
https://ijmsphr.com/index.php/ijmsphr
Model Development
The central component of this study involves the
development of deep learning models capable of
accurately predicting early-stage lung cancer.
Convolutional Neural Networks (CNNs) formed the
backbone for handling image data. A custom CNN
architecture was built from scratch consisting of
convolutional layers with ReLU activation, followed by
max-pooling layers, batch normalization, dropout
layers for regularization, and fully connected dense
layers leading to a sigmoid output node.
In addition, transfer learning models such as VGG16,
ResNet50, and InceptionV3, pre-trained on the
ImageNet dataset, were fine-tuned for lung cancer
image classification. These models offer the advantage
of learning from generalized visual features which
were further specialized by training on our lung CT
dataset.
For tabular clinical data, dense feedforward neural
networks (DNN) were constructed using fully
connected layers with appropriate activation
functions. In a hybrid setting, both CNN and DNN
branches were trained simultaneously, and their
outputs were concatenated in a fusion layer that
merged image and clinical insights before final
classification.
The models were trained using the Adam optimizer,
with binary cross-entropy as the loss function suitable
for binary classification tasks. Hyperparameters such
as learning rate, batch size, and number of epochs
were fine-tuned using grid search and validation loss
monitoring. To prevent overfitting, early stopping and
dropout regularization techniques were employed.
The entire training process was conducted on GPU-
enabled
environments
to
leverage
faster
computations.
Model Evaluation
To ensure the developed model is both accurate and
generalizable, an extensive evaluation strategy was
implemented. The dataset was split into training
(70%), validation (15%), and testing (15%) subsets
using stratified sampling to maintain the balance of
malignant and benign cases across all sets.
K-fold cross-validation with k=5 was used to evaluate
model robustness and reduce variance due to random
train-test splits. Each fold served as a temporary test set
while the others were used for training, allowing the
model to be evaluated across multiple iterations.
The primary performance metrics used included
accuracy, precision, recall, F1-score, and area under the
Receiver Operating Characteristic curve (ROC-AUC).
These metrics were chosen to reflect not only overall
model correctness but also its ability to handle class
imbalance, which is often present in medical datasets.
Confusion matrices were generated to visualize the
distribution of true positives, true negatives, false
positives, and false negatives.
Additionally, precision-recall curves were plotted to
better understand model performance on imbalanced
datasets. Comparisons among various models (custom
CNN, transfer learning models, hybrid models) were
carried out to identify the most effective architecture.
The best-performing model was selected based on the
highest ROC-AUC score and lowest false negative rate,
which is critical in cancer detection.
RESULTS
The evaluation of deep learning models for the early-
stage prediction of lung cancer was conducted through
a rigorous experimental framework that involved
stratified data splitting, k-fold cross-validation, and
extensive metric analysis. This section outlines the
comprehensive performance of five models: a custom-
designed CNN, three state-of-the-art transfer learning
architectures (VGG16, ResNet50, InceptionV3), and a
hybrid CNN-DNN model that combines both imaging
and structured clinical data.
Quantitative Evaluation of Model Performance
All models were evaluated based on five key metrics:
Accuracy, Precision, Recall (Sensitivity), F1-Score, and
Area Under the Receiver Operating Characteristic Curve
(AUC-ROC). These metrics are crucial in medical
diagnostics, where false negatives (missed cancer cases)
and false positives (incorrect cancer prediction) carry
significant implications.
pg. 68
Table 4.1: Comparative Performance Metrics of Deep Learning Models
Model
Accuracy (%)
Precision
Recall (Sensitivity)
F1-Score
AUC-ROC
Custom CNN
89.4
0.87
0.88
0.875
0.915
VGG16 (Fine-Tuned)
91.8
0.90
0.92
0.91
0.933
InceptionV3
92.5
0.91
0.92
0.915
0.939
ResNet50
93.2
0.92
0.93
0.925
0.947
Hybrid CNN-DNN
95.6
0.94
0.96
0.95
0.963
Chart 1: Evaluation of different deep learning algorithm
The results demonstrate the efficacy of deep learning
in classifying early-stage lung cancer with high
accuracy and robustness. Among all evaluated models,
the hybrid CNN-DNN architecture achieved the most
superior performance across all key metrics. The
hybrid model achieved an overall accuracy of
95.6%
, a
precision of
0.94
, a recall of
0.96
, and an AUC-ROC of
0.963
—
indicating not only its reliability in detecting
true cancer cases but also its ability to minimize false
alarms.
The custom CNN model, although effective, showed
limitations due to its simpler architecture and lack of
pretrained knowledge. It reached an accuracy of
89.4%, highlighting its potential as a baseline but
indicating the benefits of leveraging deeper or
pretrained networks.
Transfer learning models such as VGG16, ResNet50, and
InceptionV3 showed marked improvements, benefiting
from pretrained weights learned from the ImageNet
dataset. These models were fine-tuned on the lung
cancer dataset and demonstrated enhanced ability to
generalize over complex imaging features. Among
them,
ResNet50
performed best, with an accuracy of
93.2%
and AUC-ROC of
0.947
, owing to its use of
residual connections
, which help mitigate vanishing
gradient issues and facilitate deeper learning.
Notably, the
Hybrid CNN-DNN model
exhibited clear
89.4
91.8
92.5
93.2
95.6
0.87
0.9
0.91
0.92
0.94
0.88
0.92
0.92
0.93
0.96
0.87
5
0.91
0.91
5
0.92
5
0.95
0.915
0.93
3
0.93
9
0.94
7
0.96
3
C U S T O M C N N
V G G 1 6 ( F I N E -
T U N E D )
I N C E P T I O N V 3
R E S N E T 5 0
H Y B R I D C N N - D N N
MODEL PERFORMANCE
Accuracy (%)
Precision
Recall (Sensitivity)
F1-Score
AUC-ROC
International Journal of Medical Science and Public Health Research
69
https://ijmsphr.com/index.php/ijmsphr
superiority. Unlike the image-only models, it
effectively integrated
non-imaging clinical data
—
including demographic information, smoking history,
family history of cancer, and comorbidities
—
with the
imaging features learned from the CT scan data. This
fusion of heterogeneous data streams allowed the
model to extract both spatial and contextual patterns,
improving its decision-making capability. The high
recall score (0.96) is particularly important in the
clinical context, as it indicates the model’s capacity to
correctly identify almost all true positive cases
,
minimizing missed diagnoses.
Comparative
Study
and
Model
Superiority
Justification
The comparative analysis confirms that
multi-modal
deep learning approaches
, which utilize both
structured and unstructured data sources, offer
significant advantages over single-stream models.
Although pretrained image classifiers demonstrated
strong standalone performance, their predictions
lacked contextual awareness of patient-specific clinical
indicators. The hybrid CNN-DNN model's ability to
simultaneously process visual cues from CT scans and
meaningful clinical risk factors created a more
comprehensive representation of the patient profile,
resulting in superior classification outcomes.
Furthermore, the
high AUC-ROC score (0.963)
for the
hybrid model confirms its robustness in distinguishing
between early-stage lung cancer patients and healthy
individuals, even across imbalanced data subsets. This
capability is crucial in real-world screening settings,
where population variance and label noise are
common challenges.
In contrast, while the transfer learning models such as
ResNet50 and InceptionV3 captured sophisticated
spatial features from imaging data, they lacked
integration with non-visual signals. As a result,
although their precision and recall were high, they
were outperformed by the hybrid model which
provided a more
clinically nuanced prediction
framework
.
Real-World
Clinical
Integration
and
Future
Application
Given its excellent performance, the hybrid CNN-DNN
model holds substantial promise for integration into
clinical workflows. Its predictive reliability suggests it
can serve as an
automated diagnostic assistant
in
primary care clinics, specialized oncology units, and
screening programs. A suggested implementation
pipeline is as follows:
•
Radiological Workflow Enhancement
: The model
can be deployed as an add-on to existing PACS
systems to automatically screen CT scans for early
indicators of malignancy. This would assist
radiologists by highlighting at-risk patients for
further investigation.
•
EHR Integration for Personalized Risk Scores
:
Through integration with hospital EHR systems, the
model can access structured patient data in real-
time to generate individualized cancer risk profiles
that factor in non-imaging variables such as age,
lifestyle, and comorbidity burden.
•
Decision Support in Triage and Referrals
: For
patients flagged by the model as high-risk, alerts can
be generated to expedite referrals to oncology
specialists or recommend immediate diagnostic
follow-ups such as biopsies.
•
Screening Optimization for High-Risk Groups
: The
hybrid model can be adapted for use in community-
level screening programs focused on populations at
elevated risk (e.g., smokers over age 55), enabling
early detection and timely intervention.
Future Scope and Research Directions
Although the model performs well, several future
research directions remain promising. These include:
•
Incorporating genomic data or biomarkers to
further personalize predictions.
•
Extending the model for multi-class classification
(e.g., benign vs. malignant, stage I vs. II).
•
Applying federated learning frameworks to enable
secure model training across hospitals without data
centralization.
•
Validating model performance across diverse
ethnic, demographic, and geographical patient
cohorts to ensure fairness and generalizability.
DISCUSSION
The development and evaluation of deep learning
International Journal of Medical Science and Public Health Research
70
https://ijmsphr.com/index.php/ijmsphr
models for early-stage lung cancer prediction in this
study underscore the transformative potential of
artificial intelligence in oncological diagnostics. The
experimental results highlight that deep learning,
particularly convolutional neural networks (CNNs), can
effectively process complex imaging data and, when
combined with clinical features, substantially improve
predictive accuracy. Among the tested models, the
hybrid CNN-DNN framework demonstrated superior
performance in terms of precision, recall, F1-score, and
area under the ROC curve (AUC), outperforming
traditional standalone models such as simple CNNs or
dense networks.
This enhanced performance can be attributed to the
hybrid model’s ability to capture spatial patterns in
radiographic data and contextualize them with clinical
information
—
such as age, smoking history, and
familial predisposition
—
thus enabling a more
personalized and biologically plausible diagnosis. The
integration of multimodal data allows the system to
make more robust predictions, especially for
borderline or ambiguous cases where image features
alone may be insufficient.
The comparative evaluation further reveals the
limitations of traditional machine learning algorithms
when dealing with high-dimensional and unstructured
image data. Unlike conventional models, deep learning
frameworks can automatically learn hierarchical
representations from raw input without the need for
extensive feature engineering. This capability is critical
in lung cancer diagnosis, where subtle textural and
morphological variations in pulmonary nodules can
indicate malignancy but are often missed in manual or
rule-based approaches.
Moreover, the study underscores the practical viability
of applying these models in clinical settings. The high
specificity and sensitivity observed indicate that false
positives and negatives can be minimized, reducing
unnecessary interventions and enabling timely
treatment. This capability is especially important in
population-wide screening programs, where early
detection translates into higher survival rates and
lower treatment costs.
Despite these promising outcomes, the study is not
without limitations. First, while the model was trained
and validated on a well-annotated and balanced
dataset, real-world clinical data may present greater
variability in image quality, annotation consistency, and
patient demographics. Additionally, the lack of external
validation on datasets from multiple institutions could
pose challenges to generalizability. Further studies
involving
federated
learning
or
multi-center
collaborations are necessary to ensure robustness
across diverse clinical environments.
The explainability of deep learning models remains
another important concern. Although heatmaps and
Grad-CAM techniques were employed to visualize
regions of interest, ensuring clinician trust and
transparency still requires ongoing research into
interpretable AI models. Incorporating domain
knowledge and user-friendly interfaces will be essential
for clinical adoption.
From a translational standpoint, the proposed model
offers a strong foundation for integration into clinical
decision support systems (CDSS). By embedding the
model into radiology workflows or screening platforms,
clinicians can receive second-opinion assessments in
real time, assisting in the early detection and triage of
suspicious
cases.
Furthermore,
mobile
health
applications could also leverage lightweight versions of
the model to facilitate preliminary diagnostics in
underserved or rural areas.
CONCLUSION
This study successfully demonstrates that deep learning
models, particularly hybrid CNN-DNN architectures, are
highly effective in predicting early-stage lung cancer by
leveraging both imaging and clinical data. The superior
performance of the hybrid model over conventional
deep learning and machine learning models highlights
the value of multimodal data integration for accurate
and timely diagnosis.
By automating lung cancer detection, such models can
significantly reduce diagnostic delays, enhance
radiological accuracy, and contribute to improved
clinical outcomes. The findings advocate for the
deployment of AI-based diagnostic tools in routine
clinical practice, particularly in high-risk screening
populations. However, further validation using real-
world,
multicentric
data
and
interpretability
enhancements are critical steps toward safe and
effective implementation.
International Journal of Medical Science and Public Health Research
71
https://ijmsphr.com/index.php/ijmsphr
Ultimately, this research paves the way for a new era
of data-driven, AI-assisted precision medicine in
oncology, where early detection becomes not just an
aspiration but an achievable standard of care. With
continued interdisciplinary collaboration among data
scientists, clinicians, and policymakers, the integration
of deep learning in clinical oncology holds the potential
to revolutionize patient care and save countless lives.
Acknowledgement:
All the author contributed equally
REFERENCE
Ardila, D., Kiraly, A. P., Bharadwaj, S., Choi, B., Reicher,
J. J., Peng, L., ... & Tse, D. (2019). End-to-end lung
cancer screening with three-dimensional deep learning
on low-dose chest computed tomography.
Nature
Medicine
,
25
(6),
954
–
961.
https://doi.org/10.1038/s41591-019-0447-x
Choi, W., Han, K., Ko, E. S., Bae, J. M., Ko, S. Y., & Song,
S. E. (2021). Deep learning with multimodal fusion for
breast cancer diagnosis using mammography and
clinical data.
European Radiology
,
31
, 8435
–
8444.
https://doi.org/10.1007/s00330-021-07858-2
Esteva, A., Robicquet, A., Ramsundar, B., Kuleshov, V.,
DePristo, M., Chou, K., ... & Dean, J. (2019). A guide to
deep learning in healthcare.
Nature Medicine
,
25
(1),
24
–
https://doi.org/10.1038/s41591-018-0316-z
Shen, W., Zhou, M., Yang, F., Yang, C., & Tian, J. (2017).
Multi-crop convolutional neural networks for lung
nodule malignancy suspiciousness classification.
Pattern
Recognition
,
61
,
663
–
673.
https://doi.org/10.1016/j.patcog.2016.07.038
Wang, G., Liu, X., Li, C., Xu, J., Liu, Y., & Li, H. (2020). A
deep learning algorithm using CT images to screen for
Corona Virus Disease (COVID-19).
European Radiology
,
30
, 5156
–
5163.
https://doi.org/10.1007/s00330-020-
World
Health
Organization.
(2021).
Cancer
.
https://www.who.int/news-room/fact-
sheets/detail/cancer
Pabel, M. A. H., Bhattacharjee, B., Dey, S. K., Jamee, S.
S., Obaid, M. O., Mia, M. S., ... & Sharif, M. K. (2025).
BUSINESS
ANALYTICS
FOR
CUSTOMER
SEGMENTATION: A COMPARATIVE STUDY OF
MACHINE LEARNING ALGORITHMS IN PERSONALIZED
BANKING SERVICES.
American Research Index Library
, 1-
13.
Das, P., Pervin, T., Bhattacharjee, B., Karim, M. R.,
Sultana, N., Khan, M. S., ... & Kamruzzaman, F. N. U.
(2024). OPTIMIZING REAL-TIME DYNAMIC PRICING
STRATEGIES IN RETAIL AND E-COMMERCE USING
MACHINE LEARNING MODELS.
The American Journal of
Engineering and Technology
,
6
(12), 163-177.
Hossain, M. N., Hossain, S., Nath, A., Nath, P. C., Ayub,
M. I., Hassan, M. M., ... & Rasel, M. (2024). ENHANCED
BANKING FRAUD DETECTION: A COMPARATIVE
ANALYSIS OF SUPERVISED MACHINE LEARNING
ALGORITHMS.
American Research Index Library
, 23-35.
Uddin, A., Pabel, M. A. H., Alam, M. I., KAMRUZZAMAN,
F., Haque, M. S. U., Hosen, M. M., ... & Ghosh, S. K.
(2025). Advancing Financial Risk Prediction and Portfolio
Optimization Using Machine Learning Techniques.
The
American Journal of Management and Economics
Innovations
,
7
(01), 5-20.
Ahmed, M. P., Das, A. C., Akter, P., Mou, S. N., Tisha, S.
A., Shakil, F., ... & Ahmed, A. (2024). HARNESSING
MACHINE LEARNING MODELS FOR ACCURATE
CUSTOMER
LIFETIME
VALUE
PREDICTION:
A
COMPARATIVE
STUDY
IN
MODERN
BUSINESS
ANALYTICS.
American Research Index Library
, 06-22.
Nguyen, Q. G., Nguyen, L. H., Hosen, M. M., Rasel, M.,
Shorna, J. F., Mia, M. S., & Khan, S. I. (2025). Enhancing
Credit Risk Management with Machine Learning: A
Comparative Study of Predictive Models for Credit
Default Prediction.
The American Journal of Applied
sciences
,
7
(01), 21-30.
Bhattacharjee, B., Mou, S. N., Hossain, M. S., Rahman,
M. K., Hassan, M. M., Rahman, N., ... & Haque, M. S. U.
(2024). MACHINE LEARNING FOR COST ESTIMATION
AND FORECASTING IN BANKING: A COMPARATIVE
ANALYSIS
OF
ALGORITHMS.
Frontline
Marketing,Management and Economics Journal
,
4
(12),
66-83.
Hossain, S., Siddique, M. T., Hosen, M. M., Jamee, S. S.,
Akter, S., Akter, P., ... & Khan, M. S. (2025). Comparative
Analysis of Sentiment Analysis Models for Consumer
Feedback: Evaluating the Impact of Machine Learning
and Deep Learning Approaches on Business
International Journal of Medical Science and Public Health Research
72
https://ijmsphr.com/index.php/ijmsphr
Strategies.
Frontline Social Sciences and History
Journal
,
5
(02), 18-29.
Nath, F., Chowdhury, M. O. S., & Rhaman, M. M.
(2023). Navigating produced water sustainability in the
oil and gas sector: A Critical review of reuse challenges,
treatment
technologies,
and
prospects
ahead.
Water
,
15
(23), 4088.
Hossain, S., Siddique, M. T., Hosen, M. M., Jamee, S. S.,
Akter, S., Akter, P., ... & Khan, M. S. (2025).
Comparative Analysis of Sentiment Analysis Models for
Consumer Feedback: Evaluating the Impact of Machine
Learning and Deep Learning Approaches on Business
Strategies.
Frontline Social Sciences and History
Journal
,
5
(02), 18-29.
Nath, F., Asish, S., Debi, H. R., Chowdhury, M. O. S.,
Zamora, Z. J., & Muñoz, S. (2023, August). Predicting
hydrocarbon production behavior in heterogeneous
reservoir
utilizing
deep
learning
models.
In
Unconventional Resources Technology Conference,
13
–
15 June 2023
(pp. 506-521). Unconventional
Resources Technology Conference (URTeC).
Ahmmed, M. J., Rahman, M. M., Das, A. C., Das, P.,
Pervin, T., Afrin, S., ... & Rahman, N. (2024).
COMPARATIVE ANALYSIS OF MACHINE LEARNING
ALGORITHMS FOR BANKING FRAUD DETECTION: A
STUDY ON PERFORMANCE, PRECISION, AND REAL-
TIME APPLICATION.
American Research Index Library
,
31-44.
Al-Imran, M., Ayon, E. H., Islam, M. R., Mahmud, F.,
Akter, S., Alam, M. K., ... & Aziz, M. M. (2024).
TRANSFORMING BANKING SECURITY: THE ROLE OF
DEEP LEARNING IN FRAUD DETECTION SYSTEMS.
The
American
Journal
of
Engineering
and
Technology
,
6
(11), 20-32.
Pabel, M. A. H., Bhattacharjee, B., Dey, S. K., Jamee, S.
S., Obaid, M. O., Mia, M. S., ... & Sharif, M. K. (2025).
BUSINESS
ANALYTICS
FOR
CUSTOMER
SEGMENTATION: A COMPARATIVE STUDY OF
MACHINE LEARNING ALGORITHMS IN PERSONALIZED
BANKING SERVICES.
American Research Index Library
,
1-13.
Siddique, M. T., Jamee, S. S., Sajal, A., Mou, S. N.,
Mahin, M. R. H., Obaid, M. O., ... & Hasan, M. (2025).
Enhancing Automated Trading with Sentiment Analysis:
Leveraging Large Language Models for Stock Market
Predictions.
The American Journal of Engineering and
Technology
,
7
(03), 185-195.
Mohammad Iftekhar Ayub, Biswanath Bhattacharjee,
Pinky Akter, Mohammad Nasir Uddin, Arun Kumar
Gharami, Md Iftakhayrul Islam, Shaidul Islam Suhan, Md
Sayem Khan, & Lisa Chambugong. (2025). Deep Learning
for Real-Time Fraud Detection: Enhancing Credit Card
Security in Banking Systems.
The American Journal of
Engineering
and
Technology
,
7
(04),
141
–
150.
https://doi.org/10.37547/tajet/Volume07Issue04-19
Nguyen, A. T. P., Jewel, R. M., & Akter, A. (2025).
Comparative Analysis of Machine Learning Models for
Automated Skin Cancer Detection: Advancements in
Diagnostic Accuracy and AI Integration.
The American
Journal of Medical Sciences and Pharmaceutical
Research
,
7
(01), 15-26.
Nguyen, A. T. P., Shak, M. S., & Al-Imran, M. (2024).
ADVANCING EARLY SKIN CANCER DETECTION: A
COMPARATIVE ANALYSIS OF MACHINE LEARNING
ALGORITHMS FOR MELANOMA DIAGNOSIS USING
DERMOSCOPIC IMAGES.
International Journal of
Medical Science and Public Health Research
,
5
(12), 119-
133.
Phan, H. T. N., & Akter, A. (2025). Predicting the
Effectiveness of Laser Therapy in Periodontal Diseases
Using Machine Learning Models.
The American Journal
of
Medical
Sciences
and
Pharmaceutical
Research
,
7
(01), 27-37.
Phan, H. T. N. (2024). EARLY DETECTION OF ORAL
DISEASES USING MACHINE LEARNING: A COMPARATIVE
STUDY OF PREDICTIVE MODELS AND DIAGNOSTIC
ACCURACY.
International Journal of Medical Science and
Public Health Research
,
5
(12), 107-118.
PHAN, H. T. N., & AKTER, A. (2024). HYBRID MACHINE
LEARNING APPROACH FOR ORAL CANCER DIAGNOSIS
AND CLASSIFICATION USING HISTOPATHOLOGICAL
IMAGES.
Universal Publication Index e-Library
, 63-76.
