Authors

  • Mazharul Islam Tusher
    Department Of Computer Science, Monroe College, New Rochelle, New York, USA
  • Han Thi Ngoc Phan
    Dentist, Pham Hung Dental Center MTV Company Limited, Pham Hung Street, Binh Chanh district, Ho Chi Minh city, Vietnam
  • 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
  • Estak Ahmed
    Department of Computer Science, Monroe College, New Rochelle, New York, USA

DOI:

https://doi.org/10.71337/inlibrary.uz.ijmsphr.100954

Keywords:

Oral Cancer Early Detection Machine Learning Ensemble Model Clinical Data

Abstract

This study presents an integrated machine learning framework for the early detection of oral cancer, leveraging both clinical data and high-resolution imaging. The research compared several algorithms, including logistic regression, decision trees, random forests, support vector machines, and convolutional neural networks, culminating in an ensemble model that combined clinical indicators with imaging features. Results demonstrate that while traditional models provided moderate diagnostic accuracy, advanced techniques, particularly the ensemble model, achieved superior performance with an accuracy of 91%, sensitivity of 89%, specificity of 92%, and an AUC of 93%. These findings highlight that multimodal data integration significantly enhances early detection capabilities, offering a robust and practical solution for clinical implementation. The proposed framework not only improves diagnostic precision but also supports timely interventions that can potentially reduce the morbidity and mortality associated with late-stage oral cancer.


background image

International Journal of Medical Science and Public Health Research

7

https://ijmsphr.com/index.php/ijmsphr

TYPE

Original Research

PAGE NO.

07-15

DOI

10.37547/ijmsphr/Volume06Issue04-02


OPEN ACCESS

SUBMITED

22 February 2025

ACCEPTED

27 March 2025

PUBLISHED

08 April 2025

VOLUME

Vol.06 Issue04 2025

CITATION

Mazharul Islam Tusher, Han Thi Ngoc Phan, Arjina Akter, Md Rayhan
Hassan Mahin, & Estak Ahmed. (2025). A Machine Learning Ensemble
Approach for Early Detection of Oral Cancer: Integrating Clinical Data
and Imaging Analysis in the Public Health. International Journal of
Medical Science and Public Health Research, 6(04), 07

15.

https://doi.org/10.37547/ijmsphr/Volume06Issue04-02

COPYRIGHT

© 2025 Original content from this work may be used under the terms
of the creative commons attributes 4.0 License.

A Machine Learning
Ensemble Approach for
Early Detection of Oral
Cancer: Integrating Clinical
Data and Imaging Analysis
in the Public Health

Mazharul Islam Tusher

Department Of Computer Science, Monroe College, New Rochelle, New
York, USA

Han Thi Ngoc Phan

Dentist, Pham Hung Dental Center MTV Company Limited, Pham Hung
Street, Binh Chanh district, Ho Chi Minh city, Vietnam

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

Estak Ahmed

Department of Computer Science, Monroe College, New Rochelle, New
York, USA

Abstract:

This study presents an integrated machine

learning framework for the early detection of oral
cancer, leveraging both clinical data and high-resolution
imaging. The research compared several algorithms,
including logistic regression, decision trees, random
forests, support vector machines, and convolutional
neural networks, culminating in an ensemble model
that combined clinical indicators with imaging features.
Results demonstrate that while traditional models
provided moderate diagnostic accuracy, advanced
techniques, particularly the ensemble model, achieved
superior performance with an accuracy of 91%,
sensitivity of 89%, specificity of 92%, and an AUC of 93%.
These findings highlight that multimodal data
integration significantly enhances early detection
capabilities, offering a robust and practical solution for


background image

International Journal of Medical Science and Public Health Research

8

https://ijmsphr.com/index.php/ijmsphr

International Journal of Medical Science and Public Health Research

clinical implementation. The proposed framework not
only improves diagnostic precision but also supports
timely interventions that can potentially reduce the
morbidity and mortality associated with late-stage oral
cancer.

Keywords:

Oral Cancer, Early Detection, Machine

Learning, Ensemble Model, Clinical Data, Imaging
Analysis, Deep Learning, Diagnostic Accuracy.

Introduction:

Oral cancer remains a pressing global

health challenge, accounting for a significant
proportion of cancer-related morbidity and mortality
worldwide. Despite advances in medical technology,
early detection continues to be a critical factor that can
significantly improve patient outcomes and survival
rates (Chaudhary, Verma, & Kapoor, 2018). The
incidence of oral cancer is rising, particularly in regions
with high prevalence of risk factors such as tobacco
use, excessive alcohol consumption, and human
papillomavirus (HPV) infections. In many cases, late
diagnosis leads to aggressive treatments and
diminished quality of life, underlining the urgent need
for innovative diagnostic methods that can identify the
disease at its earliest stages.

The emergence of machine learning has opened new
avenues for transforming traditional diagnostic
processes in oncology. Recent advances in
computational power and the availability of large-
scale, high-quality datasets have enabled the
development of sophisticated algorithms capable of
detecting subtle patterns within complex biological
data. These techniques offer the potential to augment
clinical decision-making by providing rapid, accurate,
and reproducible assessments of patient risk factors
and disease markers. By integrating both clinical
parameters and imaging data, machine learning
models can capture a more comprehensive picture of
the pathophysiological processes underlying oral
cancer, thereby increasing the likelihood of early
detection (Wang, Li, & Chen, 2020).

In addition to enhancing diagnostic precision, the
application of machine learning in this context
supports the broader goals of personalized medicine.
Tailoring treatment strategies based on individual risk
profiles and tumor characteristics can lead to more
effective and less invasive interventions. Moreover,
the incorporation of advanced imaging analysis
through convolutional neural networks (CNNs) allows
for the automated identification of visual features that
may be imperceptible to the human eye. These
features include minute changes in tissue texture,
color variation, and lesion morphology, which are

critical indicators of early malignancy. The integration of
these diverse data sources not only bolsters the
diagnostic capabilities of the models but also provides a
robust framework for future research in cancer
detection (Khan, Ahmed, & Rahman, 2021).

Furthermore, the application of machine learning in oral
cancer detection offers the promise of overcoming
many of the limitations associated with traditional
diagnostic methods. Conventional techniques often rely
on invasive biopsies and subjective interpretation of
histopathological slides, which can lead to variability in
diagnosis. In contrast, machine learning algorithms offer
standardized assessments that can be continuously
refined through iterative training with new data. This
adaptability ensures that diagnostic tools remain
relevant and effective as new patterns of disease
emerge, particularly in the context of evolving risk
factors and treatment modalities.

This research aims to explore and validate various
machine learning methodologies for the early detection
of oral cancer. By systematically comparing traditional
statistical models with advanced deep learning
techniques, the study seeks to identify the most
effective strategies for integrating clinical and imaging
data. The ultimate goal is to develop a reliable, non-
invasive diagnostic tool that can be seamlessly
incorporated into clinical workflows, thereby facilitating
early intervention and improving patient prognoses.
Through rigorous evaluation and real-world testing, the
study aspires to contribute to the growing div of
evidence supporting the use of artificial intelligence in
medical diagnostics (Singh, Patel, & Rao, 2019).

Literature Review

The div of literature on the application of machine
learning to cancer diagnostics has grown rapidly, with
several studies highlighting the benefits of combining
clinical and imaging data. Early research in this domain
primarily focused on traditional statistical models and
decision trees, which, despite their interpretability,
were limited by their inability to capture complex, non-
linear relationships in the data (Garcia & Liu, 2017). Over
time, more sophisticated techniques such as random
forests and support vector machines emerged, offering
improved performance by effectively handling the non-
linear interactions among variables.

Recent studies have increasingly focused on the
utilization of deep learning, particularly convolutional
neural networks (CNNs), to process high-resolution
medical images. These studies have demonstrated that
CNNs can accurately identify and classify lesions with a
level of precision that approaches or even surpasses
human experts (Khan et al., 2021). In addition, the
combination of CNN outputs with traditional clinical


background image

International Journal of Medical Science and Public Health Research

9

https://ijmsphr.com/index.php/ijmsphr

International Journal of Medical Science and Public Health Research

predictors in ensemble models has been shown to
further enhance diagnostic performance, providing a
robust framework for early oral cancer detection.

Moreover, meta-analyses and systematic reviews have
underscored the importance of multimodal data
integration.

Researchers

argue

that

models

incorporating both clinical and imaging data
outperform those relying on a single modality. This
integrated approach allows for a more comprehensive
analysis, capturing the multifaceted nature of cancer
pathology. Such findings support the rationale behind
developing an ensemble model that synthesizes
various data types to improve overall diagnostic
accuracy (Singh, Patel, & Rao, 2019).

Despite these advancements, challenges remain in
translating these research findings into clinical
practice. Issues such as data heterogeneity, the need
for extensive annotated datasets, and ensuring model
interpretability for clinical decision-making are critical
areas that require further exploration. This literature
review highlights the progression from simple models

to advanced deep learning techniques and underscores
the potential benefits of multimodal data integration for
the early detection of oral cancer.

METHODOLOGY

Data Collection

This section provides an in-depth overview of the
dataset employed for the early detection of oral cancer
using machine learning techniques. The data were
sourced from multiple healthcare institutions with the
aim of capturing a diverse set of patient profiles and
clinical conditions. The dataset integrates both clinical
records and imaging studies to ensure a holistic view of
each case. It includes demographic details, lifestyle
information, clinical symptoms, biomarker readings,
and high-resolution oral cavity images, all of which
contribute to a multifactorial analysis of the disease.
The table below outlines the dataset attributes in detail,
describing each variable and its associated properties.

Attribute

Description

Data Type

Range / Categories

Patient ID

Unique identifier assigned to each patient
to ensure data integrity

Integer/String

Unique alphanumeric codes

Age

Age of the patient at the time of diagnosis,
critical for understanding risk

Integer

18 – 85

Gender

Biological sex of the patient, relevant for
epidemiological stratification

Categorical

Male, Female, Other

Tobacco Use

Indicator of tobacco consumption, a
known risk factor for oral cancer

Categorical

Yes, No

Alcohol
Consumption

Indicator of alcohol use, which can
contribute to risk assessment

Categorical

Yes, No

Clinical Symptoms Recorded clinical observations such as

pain, ulceration, or presence of lesions

Text/Categorical Varied descriptive terms

Imaging Data

High-resolution images of the oral cavity,
critical for visual analysis

Image files

JPEG/PNG formats

Histopathological
Grade

Grading

based

on

microscopic

examination of tissue samples

Categorical

Early,

Intermediate,

Advanced

Biomarker Levels

Quantitative values from blood tests
indicating the presence of cancer-linked
biomarkers

Numeric

Continuous

values

reflecting

biomarker

concentration

The dataset was carefully curated to minimize bias and
ensure an even distribution of early, intermediate, and

advanced cases. Special emphasis was placed on
collecting sufficient cases in the early stage to allow for


background image

International Journal of Medical Science and Public Health Research

10

https://ijmsphr.com/index.php/ijmsphr

International Journal of Medical Science and Public Health Research

the development of predictive models that could
detect subtle indicators of the disease. Metadata
accompanying the dataset also provide information
about imaging parameters, laboratory procedures, and
clinical assessment protocols, further enhancing the

dataset’s robustness.

Data Preprocessing

The data preprocessing stage was designed to
transform raw clinical and imaging data into a form
amenable to machine learning algorithms. Initially, the
dataset underwent a thorough cleaning process to
address issues such as missing values, inconsistencies,
and duplicate entries. Missing numerical values were
imputed using statistical measures such as the mean or
median, depending on the distribution of the data,
while categorical missing entries were filled using the
mode or a separate category indicating 'Unknown.' The
clinical text data, which included descriptions of
symptoms and patient histories, was processed using
natural language processing techniques. This involved
tokenization, stop-word removal, and lemmatization,
followed by the extraction of relevant keywords using
term frequency

inverse document frequency (TF-IDF)

methods.

The imaging data required specialized preprocessing
steps. Standardization techniques, including resizing to
a uniform resolution and normalization of pixel
intensity values, were applied to ensure consistency
across images acquired from different sources. Image
augmentation strategies such as rotation, flipping, and
cropping were also implemented to expand the
dataset and improve the robustness of the model
against variations in image orientation and lighting
conditions. The entire preprocessing workflow was
automated

using

data

pipelines,

ensuring

reproducibility and efficiency in the transformation of
raw data into structured input features.

Model Selection

Model selection was a critical phase where multiple
machine learning algorithms were rigorously
compared to determine their suitability for detecting
early-stage oral cancer. A combination of traditional
statistical models and modern deep learning
approaches were considered. For instance, logistic
regression and decision trees provided an initial
benchmark due to their interpretability and ease of
implementation. More complex models such as
random forests and support vector machines were
then evaluated to capture nonlinear relationships in
the clinical data. In parallel, convolutional neural
networks (CNNs) were selected for the analysis of
imaging data due to their proficiency in extracting
spatial features and patterns.

Evaluation criteria during this phase encompassed a
variety of performance metrics including accuracy,
sensitivity (recall), specificity, precision, and the area
under the receiver operating characteristic (ROC) curve.
Computational efficiency and the ability to integrate the
model into clinical workflows were also considered.
Extensive cross-validation was employed to assess the
consistency of model performance across different
subsets of the data. Ultimately, a balanced approach
was adopted that leveraged both the interpretability of
traditional models and the powerful pattern recognition
capabilities of deep learning architectures, ensuring that
the selected models could not only predict outcomes
with high accuracy but also provide insights that are
clinically meaningful.

Model Engineering

In the model engineering phase, the focus shifted to
refining and optimizing the selected models. Feature
engineering played a crucial role in enhancing model
performance by transforming raw data into more
informative representations. For clinical data, statistical
techniques were used to identify key predictors, and
new features were constructed by combining existing
variables to better capture complex relationships. For
example, interaction terms between age and lifestyle
factors (such as tobacco and alcohol use) were included
to assess their combined effect on cancer risk.

For imaging data, the engineering process involved
designing and tuning convolutional layers to capture
subtle visual cues indicative of early malignancy.
Advanced techniques such as transfer learning were
utilized by fine-tuning pre-trained CNN models on the
specific dataset, thereby accelerating the learning
process and improving performance with limited data.
Regularization methods, including dropout and L2
regularization,

were

implemented

to

prevent

overfitting. Dimensionality reduction techniques, like
principal component analysis (PCA), were also applied
to reduce the complexity of the feature space without
sacrificing critical information, ensuring that the model
remained both robust and computationally efficient.

Model Development

The development stage concentrated on training the
engineered models using the preprocessed dataset. This
involved splitting the dataset into training, validation,
and test sets to enable rigorous performance
monitoring throughout the iterative development
process. During training, hyperparameter optimization
techniques such as grid search and random search were
employed to identify the optimal configurations for
each model. In the case of CNNs, the training process
benefited from strategies like early stopping and
learning rate decay, which helped prevent overfitting


background image

International Journal of Medical Science and Public Health Research

11

https://ijmsphr.com/index.php/ijmsphr

International Journal of Medical Science and Public Health Research

and ensured steady convergence towards a robust
solution.

The training process was augmented by data
augmentation techniques that expanded the variability
of the input images, thus simulating a broader range of
clinical scenarios. For the traditional models working
on structured data, ensemble methods were
sometimes applied to combine predictions from
multiple models, thereby increasing overall accuracy

and reliability. Each model’s performance

was

continuously monitored using validation metrics, and
iterative refinements were made based on error
analysis and diagnostic evaluations. The integration of
multiple data modalities

clinical records and imaging

data

required careful synchronization of training

protocols to ensure that the models could leverage the
strengths of both data types concurrently.

Model Evaluation

The final stage involved an exhaustive evaluation of the
developed models, with a focus on both statistical
performance and clinical applicability. The evaluation
process utilized a comprehensive set of metrics,
including accuracy, sensitivity, specificity, precision,
recall, and the area under the ROC curve (AUC), to
quantify model performance. Detailed confusion
matrix analyses were performed to understand the
distribution of true positives, true negatives, false
positives, and false negatives. Such analyses were
critical for identifying any systematic biases or areas
where the model might underperform, particularly in
distinguishing early-stage cancer cases from non-
cancer cases.

Furthermore, validation was conducted using an
independent test set to confirm the generalizability of
the model to new, unseen data. In addition to
quantitative metrics, qualitative assessments were
made by clinical experts who reviewed the model
outputs against established diagnostic standards. These
experts provided feedback on the clinical relevance of

the predictions, ensuring that the model’s decisions

were interpretable and actionable in a real-world
setting. The evaluation phase also included robustness
checks, such as testing the model under varying
conditions and input perturbations, to confirm that the
model maintained its predictive accuracy in diverse
scenarios. This rigorous evaluation confirmed that the
integrated approach of combining clinical data with
imaging analysis can serve as a powerful tool for the
early detection of oral cancer, potentially leading to
earlier interventions and improved patient outcomes.

RESULT

This section presents an in-depth analysis of the
developed models along with a comprehensive
comparative study of their performance across different
evaluation metrics. The detailed table below
summarizes key performance indicators including
accuracy, sensitivity, specificity, precision, F1 score, the
area under the receiver operating characteristic curve
(AUC), and computational time. These metrics provide

insights into the models’ ability to detect early

-stage

oral cancer based on either clinical or imaging data, as
well as when both data sources are combined

Table 1: Model Performance

Model Name

Data Modality Accuracy

(%)

Sensitivity
(%)

Specificity
(%)

Precision
(%)

F1
Score
(%)

AUC
(%)

Computational
Time (s)

Logistic Regression

Clinical Data

82.5

79.0

84.0

81.2

80.1

85.0

0.5

Decision Tree

Clinical Data

78.0

75.5

80.0

77.0

76.2

81.0

0.3

Random Forest

Clinical Data

86.0

83.5

87.0

85.0

84.2

88.0

1.2

Support

Vector

Machine

Clinical Data

84.0

81.0

85.0

83.0

82.0

86.0

1.5

Convolutional
Neural

Network

(CNN)

Imaging Data

88.5

86.0

89.0

87.2

86.6

90.5

3.0

Ensemble

Model

(Clinical + Imaging
Data)

Combined
Clinical

&

Imaging Data

91.0

89.0

92.0

90.0

89.5

93.0

4.0

The performance table demonstrates that traditional models such as logistic regression and decision trees are

computationally efficient but fall short in accuracy and


background image

International Journal of Medical Science and Public Health Research

12

https://ijmsphr.com/index.php/ijmsphr

International Journal of Medical Science and Public Health Research

overall predictive performance. In contrast, more
sophisticated models like random forests and support
vector machines capture nonlinear interactions better,
leading to higher accuracy and AUC values. The
convolutional neural network, optimized for imaging
data, further improved sensitivity and specificity by
effectively extracting subtle visual patterns that may
indicate early malignancy.

The ensemble model, which integrates both clinical
and imaging data, stands out by outperforming all
individual models. Its superior performance is
evidenced by its higher accuracy, sensitivity,
specificity, precision, and AUC. The computational
time, though marginally longer due to the complexity
of integrating multimodal data, remains within
acceptable limits for clinical applications. The
combination of data types allows the ensemble
approach to mitigate the shortcomings of single-
modality models and provide a more comprehensive
diagnostic insight.

In addition to the table, a detailed analysis was
conducted to understand the distribution of true
positives, true negatives, false positives, and false
negatives through confusion matrix assessments. The
ensemble model, in particular, showed a balanced
distribution, minimizing false negatives

a critical

factor in early cancer detection. Robustness tests were
performed by perturbing input data and running cross-
validation cycles, which confirmed that the ensemble
model maintained consistent performance even under
varying conditions.

Comparative Study

A rigorous comparative study was performed to assess
the performance trade-offs between the models.
While traditional algorithms like logistic regression and
decision trees offer simplicity and ease of
interpretability, they are limited in capturing complex
nonlinear relationships inherent in the data. Random
forests and support vector machines improved upon
these limitations by handling nonlinearity more
effectively, but they still struggled to fully harness the
rich information provided by imaging data.

The convolutional neural network demonstrated
outstanding performance in processing imaging data,
capturing

intricate

patterns

that

are

often

imperceptible to the human eye. However, when used
in isolation, its effectiveness was slightly curtailed by
the absence of contextual clinical data. The ensemble
model, which fuses the outputs of both clinical and
imaging-based models, achieved the highest overall
performance. Its ability to integrate heterogeneous
data sources allowed it to provide a more nuanced
analysis, leading to improved diagnostic precision and

reliability. This model's enhanced sensitivity ensures
that early-stage cancers are less likely to be missed,
which is paramount in clinical settings where early
intervention

can

significantly

improve

patient

outcomes.

Real World Implementation

The real-world applicability of the ensemble model is a
critical aspect of this study. In a clinical setting, this
model can be deployed as part of an integrated
diagnostic system that assists healthcare professionals
in the early detection of oral cancer. The model works
by initially processing patient clinical data

such as

demographic information, lifestyle factors, and
laboratory

results

through

traditional machine

learning pipelines. In parallel, high-resolution images of

the patient’s oral cavity are analyzed using a

convolutional neural network. The outputs from these
separate analyses are then fused into the ensemble
model, which produces a final diagnostic prediction.

In practice, the system is designed to operate
seamlessly within hospital information systems. When a
patient is suspected of having oral cancer, their clinical
records and imaging data are automatically fed into the
model. The ensemble system generates an output that
includes a risk score along with a detailed breakdown of
the contributing factors, which aids clinicians in
understanding the underlying rationale behind the
prediction. This transparency is vital for clinical
acceptance, as it allows healthcare providers to
corroborate the model's findings with their own
expertise.

Moreover, the system is equipped with real-time
feedback mechanisms and periodic model retraining
capabilities to ensure that it stays current with emerging
patterns in patient data. It supports a decision support
framework where clinicians can override or further

investigate the model’s recommendations, thus acting

as a second pair of eyes rather than a definitive
diagnostic tool. In environments with limited access to
specialized oncological expertise, such an AI-driven
system can provide valuable preliminary assessments,
thereby streamlining the diagnostic workflow and
potentially leading to earlier interventions.

The real-world deployment also addresses key concerns
such as data privacy, system scalability, and integration
with existing electronic health records (EHRs). Rigorous
validation protocols and regulatory compliance
measures are implemented to ensure that the system
meets clinical standards and patient confidentiality
requirements. Pilot studies in several healthcare
institutions have shown promising results, with the
ensemble model demonstrating not only high diagnostic
accuracy but also significant improvements in workflow


background image

International Journal of Medical Science and Public Health Research

13

https://ijmsphr.com/index.php/ijmsphr

International Journal of Medical Science and Public Health Research

efficiency and patient management.

In conclusion, the ensemble model represents a robust
and practical solution for the early detection of oral
cancer. Its superior performance in both controlled
experiments and real-world pilot implementations
underscores its potential to become an integral part of
clinical diagnostics, ultimately contributing to
improved patient outcomes through earlier and more
accurate detection.

DISCUSSION

The results of this study underscore the significant
potential

of

integrating

machine

learning

methodologies for the early detection of oral cancer.
The comparative analysis revealed that while
traditional models like logistic regression and decision
trees offer the benefits of simplicity and computational
efficiency, they fall short in capturing the intricate non-
linear patterns inherent in complex clinical and
imaging data. Advanced models, such as random
forests, support vector machines, and especially
convolutional neural networks (CNNs), demonstrated
improved performance by effectively identifying subtle
visual and statistical cues that are indicative of early-
stage malignancy (Khan, Ahmed, & Rahman, 2021;
Wang, Li, & Chen, 2020).

The ensemble model, which synergistically combines
clinical data with imaging features, emerged as the
most robust approach. Its superior accuracy,
sensitivity,

specificity,

and

overall

diagnostic

performance suggest that the fusion of heterogeneous
data sources can substantially mitigate the limitations
of single-modality systems. This finding aligns with
previous literature that highlights the advantages of
multimodal integration in cancer diagnosis (Singh,
Patel, & Rao, 2019). Furthermore, the model's
performance across diverse test conditions and its
ability to maintain consistent predictive power even
under data perturbation emphasize its potential
applicability in dynamic clinical environments.

Despite these promising results, several challenges
remain. One primary concern is the need for extensive,
high-quality, and well-annotated datasets to train and
validate these models robustly. Data heterogeneity,
particularly in imaging modalities and clinical record
formats, may introduce variability that can affect
model performance. Additionally, while the ensemble
model provides enhanced accuracy, the increased
computational time associated with processing and
integrating multiple data types could be a limitation in
settings with constrained resources. Future research
should focus on refining data preprocessing
techniques,

exploring

more

efficient

model

architectures, and ensuring that the models remain
interpretable to clinicians. Integrating real-time
feedback and continuous learning mechanisms may also
help in adapting the system to evolving clinical practices
and patient demographics.

CONCLUSION

This study demonstrates that machine learning
techniques, particularly when leveraging an ensemble
approach that combines clinical data with imaging
analysis, can significantly enhance the early detection of
oral cancer. By comparing traditional statistical models
with advanced deep learning architectures, the research
highlights the evolution from basic diagnostic tools to
sophisticated systems capable of providing reliable,
non-invasive predictions. The ensemble model not only
outperformed individual models in terms of accuracy,
sensitivity, specificity, and AUC but also offered a robust
framework that can be integrated into clinical
workflows, facilitating timely and effective intervention.

The findings suggest that adopting such multimodal
diagnostic systems in clinical settings could improve
patient outcomes by enabling early detection and
personalized treatment planning. However, further
research is warranted to address existing challenges,
such as data variability and model interpretability, and
to optimize the balance between computational
complexity and diagnostic accuracy. Ultimately, the
integration of machine learning into oral cancer
diagnostics represents a promising step toward more
proactive and precise healthcare, offering a pathway to
reduce the mortality and morbidity associated with late-
stage diagnosis (Chaudhary, Verma, & Kapoor, 2018;
Singh, Patel, & Rao, 2019).

REFERENCE

Phan, H. T. N. (2024). EARLY DETECTION OF ORAL
DISEASES USING MACHINE LEARNING: A COMPARATIVE
STUDY

OF

PREDICTIVE

MODELS

AND

DIAGNOSTICACCURACY.

International

Journal

of

Medical Science and Public Health Research

,

5

(12), 107-

118.

Chaudhary, R., Verma, P., & Kapoor, S. (2018). Early
detection of oral cancer using machine learning: A
systematic review.

Journal of Medical Systems, 42

(4),

65

78.

Garcia, M., & Liu, S. (2017). Decision trees in cancer
diagnostics: Strengths and limitations.

Oncology

Reviews, 11

(2), 123

131.

Khan, A., Ahmed, F., & Rahman, M. (2021). Deep
learning approaches for detecting oral cancer from


background image

International Journal of Medical Science and Public Health Research

14

https://ijmsphr.com/index.php/ijmsphr

International Journal of Medical Science and Public Health Research

imaging data.

IEEE Transactions on Medical Imaging,

40

(3), 720

728.

Singh, D., Patel, R., & Rao, K. (2019). Multimodal data
integration for improved cancer diagnosis: A review.

International Journal of Cancer Research, 15

(1), 35

47.

Wang, L., Li, Y., & Chen, X. (2020). Advances in machine
learning for early cancer detection.

Artificial

Intelligence in Medicine, 104

, 101

110.

Rahman, M. M., Akhi, S. S., Hossain, S., Ayub, M. I.,
Siddique, M. T., Nath, A., ... & Hassan, M. M. (2024).
EVALUATING MACHINE LEARNING MODELS FOR
OPTIMAL CUSTOMER SEGMENTATION IN BANKING: A
COMPARATIVE STUDY.

The American Journal of

Engineering and Technology

,

6

(12), 68-83.

Akhi, S. S., Shakil, F., Dey, S. K., Tusher, M. I.,
Kamruzzaman, F., Jamee, S. S., ... & Rahman, N. (2025).
Enhancing Banking Cybersecurity: An Ensemble-Based
Predictive Machine Learning Approach.

The American

Journal of Engineering and Technology

,

7

(03), 88-97.

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.

Rishad, S. S. I., Shakil, F., Tisha, S. A., Afrin, S., Hassan,
M. M., Choudhury, M. Z. M. E., & Rahman, N. (2025).
LEVERAGING AI AND MACHINE LEARNING FOR
PREDICTING,

DETECTING,

AND

MITIGATING

CYBERSECURITY THREATS: A COMPARATIVE STUDY OF
ADVANCED MODELS.

American Research Index Library

,

6-25.

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.

Md Risalat Hossain Ontor, Asif Iqbal, Emon Ahmed,
Tanvirahmedshuvo, & Ashequr Rahman. (2024).
LEVERAGING DIGITAL TRANSFORMATION AND SOCIAL
MEDIA ANALYTICS FOR OPTIMIZING US FASHION

BRANDS’ PERFORMANCE: A MACHINE LEARNING

APPROACH.

International Journal of Computer Science

&

Information

System

,

9

(11),

45

56.

https://doi.org/10.55640/ijcsis/Volume09Issue11-05

Rahman, A., Iqbal, A., Ahmed, E., & Ontor, M. R. H.
(2024). PRIVACY-PRESERVING MACHINE LEARNING:
TECHNIQUES, CHALLENGES, AND FUTURE DIRECTIONS
IN

SAFEGUARDING

PERSONAL

DATA

MANAGEMENT.

International journal of business and

management sciences

,

4

(12), 18-32.

Iqbal, A., Ahmed, E., Rahman, A., & Ontor, M. R. H.
(2024). ENHANCING FRAUD DETECTION AND ANOMALY
DETECTION IN RETAIL BANKING USING GENERATIVE AI
AND MACHINE LEARNING MODELS.

The American

Journal of Engineering and Technology

,

6

(11), 78-91.

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
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


background image

International Journal of Medical Science and Public Health Research

15

https://ijmsphr.com/index.php/ijmsphr

International Journal of Medical Science and Public Health Research

Feedback: Evaluating the Impact of Machine Learning
and Deep Learning Approaches on Business
Strategies.

Frontline Social Sciences and History

Journal

,

5

(02), 18-29.

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.

Shakil, F., Afrin, S., Al Mamun, A., Alam, M. K., Hasan,
M. T., Vansiya, J., & Chandi, A. (2025). HYBRID MULTI-
MODAL DETECTION FRAMEWORK FOR ADVANCED
PERSISTENT THREATS IN CORPORATE NETWORKS
USING

MACHINE

LEARNING

AND

DEEP

LEARNING.

American Research Index Library

, 6-20.

Rishad, S. S. I., Shakil, F., Tisha, S. A., Afrin, S., Hassan,
M. M., Choudhury, M. Z. M. E., & Rahman, N. (2025).
LEVERAGING AI AND MACHINE LEARNING FOR
PREDICTING,

DETECTING,

AND

MITIGATING

CYBERSECURITY THREATS: A COMPARATIVE STUDY OF
ADVANCED MODELS.

American Research Index Library

,

6-25.

Das, A. C., Rishad, S. S. I., Akter, P., Tisha, S. A., Afrin, S.,
Shakil, F., ... & Rahman, M. M. (2024). ENHANCING
BLOCKCHAIN SECURITY WITH MACHINE LEARNING: A
COMPREHENSIVE STUDY OF ALGORITHMS AND
APPLICATIONS.

The American Journal of Engineering

and Technology

,

6

(12), 150-162.

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.

Akhi, S. S., Shakil, F., Dey, S. K., Tusher, M. I.,
Kamruzzaman, F., Jamee, S. S., ... & Rahman, N. (2025).
Enhancing Banking Cybersecurity: An Ensemble-Based
Predictive Machine Learning Approach.

The American

Journal of Engineering and Technology

,

7

(03), 88-97.

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.

References

Phan, H. T. N. (2024). EARLY DETECTION OF ORAL DISEASES USING MACHINE LEARNING: A COMPARATIVE STUDY OF PREDICTIVE MODELS AND DIAGNOSTICACCURACY. International Journal of Medical Science and Public Health Research, 5(12), 107-118.

Chaudhary, R., Verma, P., & Kapoor, S. (2018). Early detection of oral cancer using machine learning: A systematic review. Journal of Medical Systems, 42(4), 65–78.

Garcia, M., & Liu, S. (2017). Decision trees in cancer diagnostics: Strengths and limitations. Oncology Reviews, 11(2), 123–131.

Khan, A., Ahmed, F., & Rahman, M. (2021). Deep learning approaches for detecting oral cancer from imaging data. IEEE Transactions on Medical Imaging, 40(3), 720–728.

Singh, D., Patel, R., & Rao, K. (2019). Multimodal data integration for improved cancer diagnosis: A review. International Journal of Cancer Research, 15(1), 35–47.

Wang, L., Li, Y., & Chen, X. (2020). Advances in machine learning for early cancer detection. Artificial Intelligence in Medicine, 104, 101–110.

Rahman, M. M., Akhi, S. S., Hossain, S., Ayub, M. I., Siddique, M. T., Nath, A., ... & Hassan, M. M. (2024). EVALUATING MACHINE LEARNING MODELS FOR OPTIMAL CUSTOMER SEGMENTATION IN BANKING: A COMPARATIVE STUDY. The American Journal of Engineering and Technology, 6(12), 68-83.

Akhi, S. S., Shakil, F., Dey, S. K., Tusher, M. I., Kamruzzaman, F., Jamee, S. S., ... & Rahman, N. (2025). Enhancing Banking Cybersecurity: An Ensemble-Based Predictive Machine Learning Approach. The American Journal of Engineering and Technology, 7(03), 88-97.

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.

Rishad, S. S. I., Shakil, F., Tisha, S. A., Afrin, S., Hassan, M. M., Choudhury, M. Z. M. E., & Rahman, N. (2025). LEVERAGING AI AND MACHINE LEARNING FOR PREDICTING, DETECTING, AND MITIGATING CYBERSECURITY THREATS: A COMPARATIVE STUDY OF ADVANCED MODELS. American Research Index Library, 6-25.

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.

Md Risalat Hossain Ontor, Asif Iqbal, Emon Ahmed, Tanvirahmedshuvo, & Ashequr Rahman. (2024). LEVERAGING DIGITAL TRANSFORMATION AND SOCIAL MEDIA ANALYTICS FOR OPTIMIZING US FASHION BRANDS’ PERFORMANCE: A MACHINE LEARNING APPROACH. International Journal of Computer Science & Information System, 9(11), 45–56. https://doi.org/10.55640/ijcsis/Volume09Issue11-05

Rahman, A., Iqbal, A., Ahmed, E., & Ontor, M. R. H. (2024). PRIVACY-PRESERVING MACHINE LEARNING: TECHNIQUES, CHALLENGES, AND FUTURE DIRECTIONS IN SAFEGUARDING PERSONAL DATA MANAGEMENT. International journal of business and management sciences, 4(12), 18-32.

Iqbal, A., Ahmed, E., Rahman, A., & Ontor, M. R. H. (2024). ENHANCING FRAUD DETECTION AND ANOMALY DETECTION IN RETAIL BANKING USING GENERATIVE AI AND MACHINE LEARNING MODELS. The American Journal of Engineering and Technology, 6(11), 78-91.

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 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.

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.

Shakil, F., Afrin, S., Al Mamun, A., Alam, M. K., Hasan, M. T., Vansiya, J., & Chandi, A. (2025). HYBRID MULTI-MODAL DETECTION FRAMEWORK FOR ADVANCED PERSISTENT THREATS IN CORPORATE NETWORKS USING MACHINE LEARNING AND DEEP LEARNING. American Research Index Library, 6-20.

Rishad, S. S. I., Shakil, F., Tisha, S. A., Afrin, S., Hassan, M. M., Choudhury, M. Z. M. E., & Rahman, N. (2025). LEVERAGING AI AND MACHINE LEARNING FOR PREDICTING, DETECTING, AND MITIGATING CYBERSECURITY THREATS: A COMPARATIVE STUDY OF ADVANCED MODELS. American Research Index Library, 6-25.

Das, A. C., Rishad, S. S. I., Akter, P., Tisha, S. A., Afrin, S., Shakil, F., ... & Rahman, M. M. (2024). ENHANCING BLOCKCHAIN SECURITY WITH MACHINE LEARNING: A COMPREHENSIVE STUDY OF ALGORITHMS AND APPLICATIONS. The American Journal of Engineering and Technology, 6(12), 150-162.

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.

Akhi, S. S., Shakil, F., Dey, S. K., Tusher, M. I., Kamruzzaman, F., Jamee, S. S., ... & Rahman, N. (2025). Enhancing Banking Cybersecurity: An Ensemble-Based Predictive Machine Learning Approach. The American Journal of Engineering and Technology, 7(03), 88-97.

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.

Most read articles by the same author(s)

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, 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: Vol. 6 No. 05 (2025)

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, 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: Vol. 6 No. 05 (2025)