Smt.

Indira Gandhi College of Engineering

Affiliated to University of Mumbai
Department of CSE(AIML)
(2024–2025)

LUNG CANCER DETECTION AND

CLASSIFICATION

1. Hasan Patel

2. Sandesh Patil

3. Jay Pardeshi

GUIDE

PROF. VENKAT PATIL

 INDEX

1. ABSTRACT

2. INTRODUCTION

3. LITERATURE REVIEW

4. METHODOLOGY

5. RESULTS

6. CONCLUSION
 ABSTRACT

Lung cancer remains a significant public health challenge, being one

of the leading causes of cancer-related mortality worldwide. Early
detection is crucial for improving patient outcomes and survival
rates. This project aims to develop an automated lung cancer
detection and classification system using deep learning techniques,
specifically convolutional neural networks (CNNs). A dataset of
lung CT scans was utilized to train the model, which distinguishes
between cancerous and non-cancerous images. Through data
augmentation and the application of a pre-trained model, the system
achieved high accuracy in classification tasks. Evaluation metrics
such as accuracy, precision, recall, and F1-score were employed to
assess model performance. The results indicate that the proposed
method can significantly aid in early diagnosis and provide a
reliable tool for healthcare professionals, ultimately contributing to
better patient management and treatment strategies.
 Introduction

Background: Lung cancer is one of the most prevalent and deadly

forms of cancer globally, accounting for a significant number of
cancer-related deaths each year. The disease often goes undetected
until it reaches advanced stages, leading to poor prognosis and
limited treatment options. Traditional diagnostic methods, such as X-
rays and biopsies, can be invasive and may lack the sensitivity
required for early detection. As a result, there is a pressing need for
more effective and reliable diagnostic tools.

Objective: This project aims to develop a machine learning-based

system for the detection and classification of lung cancer using deep
learning techniques. By leveraging convolutional neural networks
(CNNs), the project seeks to automate the analysis of lung CT scans,
facilitating the differentiation between cancerous and non-cancerous
images. The ultimate goal is to enhance diagnostic accuracy and
provide timely support for healthcare professionals.

Motivation: The motivation behind this project stems from the urgent
need for improved diagnostic methods in lung cancer detection. As
the medical field increasingly embraces artificial intelligence, the
potential to harness deep learning for medical image analysis
represents a significant advancement in patient care. This project
aspires to contribute to early diagnosis, ultimately improving patient
outcomes and paving the way for more effective treatment strategies.
 Literature Review

Lung cancer detection and classification have evolved significantly

over the past few decades, particularly with the advent of medical
imaging technologies and machine learning techniques. Traditional
diagnostic methods, such as chest X-rays and computed tomography
(CT) scans, have been the cornerstone of lung cancer diagnosis.
However, these methods often face challenges related to sensitivity
and specificity, leading to the need for enhanced diagnostic
approaches.

Traditional Diagnostic Methods: Historically, lung cancer

detection relied heavily on imaging techniques and invasive
procedures like biopsies. X-rays have been widely used due to their
availability and cost-effectiveness, but they are limited in
sensitivity, particularly in early-stage cancers. CT scans provide
higher resolution images and have become the standard for lung
cancer screening; however, interpreting these images can be
subjective and requires significant expertise. As a result, there is a
growing interest in leveraging AI to improve diagnostic accuracy
and reduce human error.

Machine Learning and Deep Learning in Medical Imaging:

Recent advancements in artificial intelligence have enabled the
development of machine learning models capable of analyzing
medical images with remarkable accuracy. In particular, deep learning
techniques, especially convolutional neural networks (CNNs), have
shown great promise in image classification tasks. Studies have
demonstrated that CNNs can effectively identify and classify lung
nodules in CT scans, often outperforming traditional methods. For
instance, a study by Ardila et al. (2019) utilized deep learning
algorithms to analyze chest X-rays, achieving higher accuracy than
radiologists in detecting lung cancer.
Dataset Utilization: The success of deep learning models relies
significantly on the quality and quantity of data available for training.
Several publicly available datasets, such as the LIDC-IDRI (Lung
Image Database Consortium Image Database Resource Initiative) and
NSCLC (Non-Small Cell Lung Cancer) datasets, have been
instrumental in training and validating AI models. These datasets
contain annotated images that allow researchers to develop robust
models that generalize well to new cases. The use of data
augmentation techniques has also been widely studied to enhance
model performance by artificially increasing the dataset size and
variability.

Comparative Studies: Various studies have compared the

performance of AI-driven diagnostic systems with that of human
experts. For example, a systematic review by Ghafoor et al. (2020)
highlighted that AI models could achieve performance levels
comparable to, or even exceeding, that of experienced radiologists in
detecting lung cancer from imaging data. These findings underscore
the potential of AI as a complementary tool in clinical settings,
enhancing the efficiency and accuracy of lung cancer diagnoses.

Challenges and Limitations: Despite the promising advancements,

several challenges remain in integrating AI into clinical practice.
Issues such as data privacy, the need for standardized protocols, and
the interpretability of AI models pose significant barriers.
Furthermore, the reliance on large, annotated datasets necessitates
collaboration between researchers and medical institutions to create
robust training resources.
 Methodology
The methodology for this project consists of several key steps,
including dataset selection, data preprocessing, model development,
training, and evaluation. Each of these steps is crucial for building an
effective lung cancer detection and classification system using deep
learning techniques.

1. Dataset Description

For this project, the LIDC-IDRI (Lung Image Database

Consortium Image Database Resource Initiative) dataset was
utilized. This dataset contains a comprehensive collection of
annotated lung CT scans, including images labeled for the presence of
nodules and various types of lung cancer. It consists of a diverse
range of cases, ensuring representation across different cancer stages
and types. The dataset allows for both training and testing of the deep
learning model, providing a solid foundation for developing a robust
classification system.

2. Data Preprocessing

Data preprocessing is essential to ensure the model's effectiveness and

performance. The following preprocessing steps were implemented:

 Data Cleaning: Images were inspected for any corruption or

anomalies, and invalid images were removed from the dataset.
 Image Resizing: All CT scan images were resized to a standard
dimension (e.g., 224x224 pixels) to maintain consistency and
compatibility with the CNN model input.
 Normalization: Pixel values were normalized to a range of [0,
1] to improve convergence during training. This process helps
the model learn more effectively by ensuring that input values
are scaled similarly.
  Data Augmentation: To enhance the model's generalization
capabilities, data augmentation techniques were employed. This
included random rotations, shifts, flips, and zooms to artificially
increase the dataset's size and variability. These transformations
help prevent overfitting by providing the model with diverse
training examples.

3. Model Architecture

The deep learning model developed for this project is based on a

convolutional neural network (CNN) architecture. The model consists
of several layers, including:

 Convolutional Layers: These layers extract features from the

input images by applying various filters. Multiple convolutional
layers were stacked to learn complex patterns in the images.
 Activation Functions: The ReLU (Rectified Linear Unit)
activation function was used to introduce non-linearity, allowing
the model to learn more complex representations.
 Pooling Layers: Max pooling layers were added after
convolutional layers to down-sample feature maps, reducing
dimensionality while retaining essential information.
 Fully Connected Layers: After several convolutional and
pooling layers, the output was flattened and passed through fully
connected layers, culminating in a softmax layer that provides
probabilities for each class (cancerous and non-cancerous).

4. Training Process

The model was trained using the following parameters:

 Split of Data: The dataset was divided into training, validation,

and test sets, typically using a split ratio of 70% for training,
15% for validation, and 15% for testing.
 Batch Size: A batch size of 32 was chosen to balance memory
usage and training efficiency.
 Epochs: The model was trained for a predetermined number of
epochs (e.g., 50), allowing sufficient time for the model to learn
 from the training data while monitoring performance on the
validation set.
 Loss Function: The categorical cross-entropy loss function was
used to measure the model's performance during training.
 Optimizer: The Adam optimizer was employed for efficient
training, allowing for adaptive learning rates.

5. Evaluation Metrics

To evaluate the performance of the trained model, the following

metrics were used:

 Accuracy: The proportion of correctly classified images out of

the total number of images in the test set.
 Precision: The ratio of true positive predictions to the total
predicted positives, indicating the model's ability to avoid false
positives.
 Recall: The ratio of true positive predictions to the total actual
positives, reflecting the model's ability to identify all relevant
instances.
 F1-Score: The harmonic mean of precision and recall, providing
a balance between the two metrics.

The evaluation process involved analyzing the model's performance

on the test set to ensure its generalization capability. A confusion
matrix was also generated to visualize the model's classification
performance and identify areas for improvement.
Results
 Conclusion
In this project, machine learning algorithms were successfully
implemented to detect and classify lung cancer based on clinical
data. The Random Forest and XGBoost models provided the best
results, showing that both categorical and numerical features are
crucial in identifying lung cancer patients. Further improvements
can be made by utilizing larger datasets, integrating deep learning
models, or refining the feature engineering process.This study
demonstrates that AI and ML can be effective tools in supporting
early lung cancer detection, potentially aiding in saving lives
through timely diagnosis.

Future Scope
Incorporating more sophisticated deep learning techniques like
convolutional neural networks (CNNs) for image-based detection,
combined with clinical data.Utilizing additional medical datasets to
generalize the model further and improve its
performance.Implementing a real-time prediction system integrated
into clinical workflows to assist doctors in lung cancer screening.