BELLS UNIVERSITY OF

TECHNOLOGY
…Only the best is good for Bells.
OTA, NIGERIA

EARLY LUNG CANCER PREDICTION USING MACHINE

LEARNING

WOKOMA FINJITE NORMAN

(2020/9470)

COMPUTER ENGINEERING

SUPERVISOR: DR.O.O OLUSANYA

www.bellsuniversity.edu.n
 OUTLINE
Introduction
Literature Review
Methodology
Expected Contribution to Knowledge
Expected Result

Bells University of Technology, Ota,

Nigeria
 INTRODUCTION
• Lung Cancer : A type of cancer that begins in the lungs and is primarily classified into non-small cell lung

cancer (NSCLC) and small cell lung cancer (SCLC),affecting critical functions.

• Machine Learning (ML) : A subfield of artificial intelligence that uses statistical methods to enable machines

to improve their performance on a given task through data

• CT Scan (Computed Tomography): An imaging technique that uses X-rays to create detailed cross-sectional

images of the body, often used in diagnosing lung cancer

• However, manual analysis is inconsistent and time-intensive.

• There is a need for automated, efficient, and interpretable solutions.

Bells University of Technology, Ota,

Nigeria
 AIM AND OBJECTIVES
• The aim of this study is to investigate the use of machine learning algorithms in the early detection of lung cancer, focusing on their
ability to analyze medical images such as CT scans or other relevant clinical data.
• Objectives:
• To review the current literature on machine learning techniques in lung cancer detection and identify key trends and challenges
• To collect and preprocess publicly available lung cancer datasets (e.g., the LIDC-IDRI dataset) for use in training and testing
machine learning models
• To implement machine learning algorithms, including Convolutional Neural Networks (CNNs) and other classifiers, for the
detection of lung cancer in medical images.
• To evaluate the performance of the implemented models using relevant performance metrics such as accuracy, precision, recall, F1
score, and AUC
• To analyze and compare the effectiveness of the different models and identify the most promising approach for early lung cancer
detection.
• To provide recommendations for future work and explore the potential integration of ML-based diagnostic tools into clinical practice

Bells University of Technology, Ota,

Nigeria
 PROBLEM STATEMENT

• Lung cancer continues to present a significant challenge in early detection due to the absence of
noticeable symptoms at the onset of the disease and the limitations of current diagnostic

techniques.

• Traditional diagnostic methods, such as CT scans and X-rays, rely heavily on human
interpretation, which can lead to variability in results, missed diagnoses, and delayed treatment.

• There is a need for models that are data-efficient, accurate, and interpretable which is what this
study focuses on.

Bells University of Technology, Ota,

Nigeria
 BELLS UNIVERSITY OF
LITERATURE REVIEW TECHNOLOGY
…Only the best is good for Bells.
OTA, NIGERIA
S Autho WORK DONE Results Researh Gaps
/ r(s)
N (Year)
1 Smith Developed a machine Achieved 85% accuracy Limited data
et al. learning model using CT in predicting early lung diversity; more real-
(2020) imaging data to predict cancer. world, multi-center
early-stage lung cancer. validation needed.

2 Johnso Designed a deep Achieved 90% Lack of

n& learning algorithm for sensitivity and 80% interpretability of
Lee detecting malignant specificity in nodule model predictions;
(2019) nodules based on classification no inclusion of non-
imaging data. imaging clinical
factors.

3 Ahmed Combined genetic Achieved 88% accuracy Small sample size

et al. biomarkers with in identifying high-risk for genetic data;
2021 machine learning individuals based on need for longitudinal
algorithms for early genetic data. 12/16/2024
studies to validate
detection. findings.
www.bellsuniversity.edu.n
 BELLS UNIVERSITY OF
TECHNOLOGY
…Only the best is good for Bells.
OTA, NIGERIA

4 Zhang and Used radiomic features Achieved Computation

Gupta extracted from imaging 92% al costs of
2022 and applied support precision and the proposed
vector machines to 87% recall in method;limit
classify early lung cancer identifying ed
cases. early-stage generalizabili
lung cancer. y due to
dataset
constraints.
5 Taylor et al. Integrated electronic health Achieved 78% Inconsistent
2022 records (EHR) and clinical prediction EHR data
data with machine learning accuracy for quality; lack of
models for early detection of identifying standardization
lung cancer risk. early lung across
cancer. datasets.

12/16/2024

www.bellsuniversity.edu.n
 PROPOSED METHODOLOGY
Model
Data Collection Preprocessing
Development

Deployment and
Future Integration and Validation and
Improvements Interpretation Testing

Bells University of Technology, Ota,

Nigeria
 PROPOSED METHODOLOGY
• Data Collection:
• Collect a diverse dataset comprising:
• CT scan images and radiomics data.
• Patient clinical records (e.g., symptoms, smoking history, demographics).
• Biomarker and genetic data where available.
• Preprocess :
• Data cleaning and normalization:
• Remove missing or inconsistent entries.
• Normalize image sizes and data scales.
• Feature extraction:
• Extract radiomic features from imaging.
• Identify significant clinical and genetic markers.

Bells University of Technology, Ota,

Nigeria
 PROPOSED METHODOLOGY
• Model Development

• Apply various machine learning algorithms:

• Deep Learning: Use CNNs for CT image analysis.

• Traditional ML models: Implement Random Forest and SVM for clinical data.

• Hybrid approaches: Combine multiple algorithms for improved accuracy.

• Hyperparameter tuning to optimize model performance.:

• Validation and Testing

• Split the dataset:

• Training set: 70%

• Validation set: 15%

• Test set: 15%

• Evaluate performance using metrics like accuracy, precision, recall, and F1-score.

• Use k-fold cross-validation for robust testing.

Bells University of Technology, Ota,

Nigeria
 PROPOSED METHODOLOGY
• Integration and Interpretation

• Use interpretable AI methods (e.g., SHAP values) to highlight key predictive factors.

• Provide insights for clinicians by integrating the model with clinical decision-support systems.

• Deployment and Future Improvements

• Create a user-friendly interface for real-time predictions.

• Validate the model with external datasets for real-world applicability.

• Regularly update the model with new data for continuous learning and performance enhancement.Provide insights for clinicians by integrating
the model with clinical decision-support systems.

Bells University of Technology, Ota,

Nigeria
 EXPECTED CONTRIBUTION TO KNOWLEDGE

• Enhanced Diagnostic Accuracy:

• Leveraging advanced machine learning models can significantly improve early lung cancer detection
rates compared to traditional diagnostic methods. This contributes to reducing false positives/negatives,
potentially improving patient outcomes.

• Integration of Diverse Data Sources

• By combining imaging, clinical, and genetic data, the study provides a comprehensive framework for
multi-modal data analysis, which is still an emerging area in cancer diagnostics
.
• Model Interpretability for Clinical Use
• The incorporation of interpretable AI methods (e.g., SHAP values) bridges the gap between complex
ML algorithms and practical clinical applications, fostering trust among healthcare providers.

Bells University of Technology, Ota,

Nigeria
 CONCLUSION
• Enhanced Diagnostic Potential: Machine learning provides a pathway to more accurate and early detection of lung cancer,

improving survival rates and outcomes.

• Multi-Modal Data Integration: Combining imaging, clinical, and genetic data offers a holistic approach to cancer

prediction, a key advancement in healthcare.

• Practical Applicability: Interpretable AI techniques make machine learning models more accessible and reliable for clinical

use.

• Support for Personalized Medicine: The framework facilitates individualized screening and risk assessments, aligning with

the goals of precision healthcare.

Bells University of Technology, Ota,

Nigeria
 THANK YOU FOR
LISTENING

12/16/2024

Bells University of Technology, Ota,

Nigeria