COM748 Masters Research Project

Project Proposal Form

Student Name Student Number Supervisor Name

Project Title
Brain Tumour detection using CNN machine learning

Problem Statement (100 words)

Brain tumours are a serious health challenge, disrupting essential brain functions like waste
removal and nutrient absorption, and leading to complications such as kidney disease and vascular
issues. These conditions can worsen quickly, often with limited warning, making early detection vital
for improving patient outcomes. However, reliable patient data is lacking, and current tools struggle
with consistent, early identification. This project proposes creating a deep learning-based tool to
detect and monitor brain tumours in real-time, allowing healthcare providers to catch and respond to
changes early. With this approach, we aim to offer a transformative, accessible solution for timely
diagnosis.

Background Research (500 words)

Brain tumours are abnormal cell growths within the brain or surrounding tissues. They are classified
into various types based on their origin, malignancy, and growth patterns. Gliomas, for instance, are
derived from glial cells, which support and protect neurons. These tumours constitute about 30% of
all brain and central nervous system tumours, with two main categories: , we have divided glioma
into low-grade gliomas include Grade I or II gliomas and high-grade gliomas that include Grade III or
IV gliomas. Gliomas are of low grade most of which are slow growing while the high-grade gliomas
are very malignant among which is the GBM Since it stems from the astrocytes also known as the
star-shaped glial cells that support the neurons of the brain, the type of brain tumour is the
astrocytoma. Astrocytoma is also divided in to low-grade astrocytoma and high-grade astrocytoma.
The latter comprises anaplastic astrocytoma, which usually make for highly malignant tumors and
are usually not easy to excise by operation on account of their infiltrative characteristics. Most
Primary intracranial tumours originate from the meninges and are benign but when large or situated
in critical areas are a cause of health complications and have to be removed surgically. All pituitary
tumours are usually indicative of benign tumours located at the base of the brain near the optic
chiasm and they can cause hormonal disruption and all sorts of symptoms many of which are listed
below in relation to their hormonal origin.

Imaging with special attention to MRI is used to diagnose brain tumours correctly since its structural
resolution is comparatively high. However, there are problems with distininction between malignant
and benign tumours, as well as the definition of the type of tumour. This is where getting Artificial
Intelligence or AI can play a strong role. AI and particularly Machine Learning are rapidly
transforming the way medical image is analyzed. In the last few years, research has demonstrated
the usefulness of AI in helping to identify and categorise brain tumours using MRI scans. The
application of such approaches as deep learning solves the problem of differentiation between
various types of tumours, which helps to develop the appropriate further therapeutic tactics at the
earlier stages.

The implementation of AI in healthcare is of relatively long-standing beginning with facial intelligence

as early as 1956. In an early period of AI development, research has centred on the deployment and
application of AI as a symbolic processor. However, considerable progress was made in the sixties
and the seventies when the United States Department of Défense began to sponsor work on
creating end user computers which would emulate the psychological characteristics of the human

1
brain. Since then, AI has been installed in numerous areas of applicability, one of which is the
medical field, which has put AI to use doing everything from clerical work to analyzing patients. To
date, AI has been most effective in medical imaging, especially y. It benefits the radiologists by
quickly diagnosing medical images because of the patterns and abnormalities it detects on its own.

In the case of early detection of brain tumour the application of AI algorithms especially the one
based on *deep can find out even slight changes in the MRI scans that are not conspicuous to the
human eye. These utensils have been shown efficacious in the detection of gliomas, astrocytoma
and meningiomas, and sometimes, even benign and malignant tumours. The application of AI in
medical imaging has the capability of minimising diagnostic mistakes, quicken diagnosis and in the
long run, enhance the result of the patient.

As the field evolves, the concern is given toward the specificities of the models like data
diversification and interpretability of the model in AI. Additionally, the application of AI in health
sectors keeps on growing, with higher focus on precision or personal medicine which is informed by
AI enhanced analysis customized for every user. These advancements enable this project, dedicated
to constructing Artificial Intelligence methods for MRI image-based brain tumors classification, to
advance a more precise and effective method for diagnosing and treating brain tumours.

Project Aim and Objectives (300 words)

The objective of this work is to design a deep learning solution based on Convolutional Neural
Networks (CNNs) with focus to classify and detect chronic brain tumour diseases from MRI images.
This work is planned to work with imbalanced datasets and guarantee stable accuracy even in the
case of unknown data, which would help specialists in the medical field identify and classify brain
tumors more effectively and with high accuracy.

Project Objectives:

1. Data Collection and Preprocessing:

 The main source of the MRI datasets for the current study is Kaggle, which offers
images of brain tumours.
 First, reduce the image size to singular dimensions, balance the data by oversampling
or augmentation to rectify skewed data and enhance model performance by
sharpening image quality.balances using techniques such as oversampling or
augmentation and improve image quality for better model performance.

2. Model Development:
 Develop a CNN model specific for classification of brain tumours and optimize the
design using advances in deep learning frameworks and GPU computing for
analysing volumetric images.
 Perform data preprocessing on the collected dataset for usage in training the CNN
model on different types of brain tumours – benign and malignant.

3. Model Evaluation:
 Assess the model by using classification measures which include accuracy, precision,
recall rates as well as F1 score especially for imbalanced data sets and for the model
capacity to distinguish between various tumour types.
 Tune up the model using processes such as hyperparameter optimization,
standardization, and early exit to improve its performance.

4. Testing with Unseen Data:

 Use the model on new MRI images to check if it’s functional in real life situations and
how strong it is.

5. Application Development:
 Building an intuitive interface for disease classification using the trained CNN model for
the case of brain tumours with a view of assisting clinicians in their diagnosis.

2
The goal of this project is to improve AI-assisted methods for screen and diagnose of brain
tumours, thus increasing the quality of patients’ treatment.

Methodology/approach (400 words)

Methodology/Approach

Methodology including data collection, preprocessing, model design, training, and evaluation, using
established methods in the field of medical imaging and machine learning.

1. Data Collection and Preprocessing

The first step is to obtain MRI scan datasets thus comprising of labelled images of brains with
tumours. These datasets have different forms of tumours including gliomas, astrocytoma, and
meningiomas and are usually obtained from open-access fields such as Kaggle. To avoid
overfitting, and work with fewer samples, data enlargement methods like image rotation, flipping,
and zooming are used to enhance the inspired dataset.

This entails making images black and white, enhancing image contrast to improve visibility and
lastly binarization to bring out tumour characteristics. That is, if there is any noise in the images,
filtering techniques are applied in order to enhance the input data most of the time.

2. Model Design and Development

A Convolutional Neural Network (CNN) is employed as the foundational model architecture due to
its suitability in Image Recognition. The CNN model for the proposed system has multiple
convolutional layers to extract the features of the images, pool layers to reduce dimensions and fully
connected layers for classification. The model shall use activation function like ReLU to introduce
non-linearity to favors learning in the model. For the adaptations of the hyperparameters the
learning rate, the batch size and the number of epochs will be adjusted.

3. Training and Validation

Supervised learning paradigm is used for the training of the model because the labeled data is
employed to show the model how to identify brain tumors. To ensure that over fitting is prevented,
the model’s performance is tested using the cross-validation techniques that makes the model to
generalize on unseen data. Besides, the dataset is divided into training, validation and test to get
the desired accuracy of the model and to check its reliability.

4. Evaluation and Testing

After the model has received training, the quality of the method is evaluated via three common
parameters: precision, recall, F1-score and accuracy. On a different data set – the test data set
which consist of brain scans that the model never encountered before, its performance in making
predictions is evaluated. The confusion matrix is also created in order to plot the performance and
determine if there is any misclassification.

5. Software and Hardware

On the CNN model, basic programing tool is Python alongside the complementary deep learning
libraries such as TensorFlow and Keras. These frameworks allow the creation of the operation,
training, and assessing capabilities of the neural network. The training process is computationally
expensive and thus the system is implemented on hardware with adequate GPU to handle model

3
 training.

Legal/ Social/ Ethical Issues (200 words)

To maintain professionalism in this research project, several legal, social and ethical factors that
need to be looked at so that it can conform to the legal, social, and ethical regulations of the
profession in computer science.

Legal Considerations:

This project applies medical image datasets which mainly include MRI outcomes; these datasets
are usually associated with subject’s health information. In particular, the used datasets must be
public, anonymized and do not violate the relevant legislation, including GDPR in the EU or HIPAA
in the United States. The research also cannot use any patient identifiable data while conducting
the research without prior consent, something that requires that data handling and storage
processes are legal and secure.

Social Considerations:

The research study would be relevant to healthcare systems because it makes it possible to
accelerate the chances of detecting brain tumours accurately. Nonetheless, the application of such
technology within clinical practice contexts must also be mindful of other social possibilities, which
include availability and discriminant of the model. However, if more variety is not found in the
dataset, then it becomes a problem of imprecise forecasting for such subgroups. This has to be
answered using balanced and comprehensive data set.

Ethical Considerations:

Ethically, the project must prioritize the well-being of individuals who might be impacted by the
results. This includes ensuring that the AI system does not provide harmful or incorrect diagnoses.
Ethical approval will be sought for the use of medical data, and the project will follow the ACM Code
of Ethics, ensuring transparency, fairness, and accountability in the research process. All
participants involved, even indirectly, will be respected, and data privacy will be prioritized.

Project & Risk Management (300 words)

Managing risks is a crucial aspect of ensuring the success of the brain tumor detection project
using deep learning techniques. Three primary types of risks must be considered: These
categories occur in the form of health and safety risks, management risks, and technical risks.
1) Health and Safety Risks:
This project is more of a computational project, and as such does not involve any potential
physical hazards of electrical equipment or environments. The most relevant issue in relation
to health risks is the risk of stress originating from hours of sitting in front of a computer. To
avoid this imbalance, sufficient intervals of rest will be incorporated and optimized workspace
design will also be incorporated to eliminate the possibility of RSI’s.
2) Management Risks:
There is a lot of management risks, but one of them is to implement the project on time and
provide all the necessary quality standards at the same time. If the software is developed too
late, it may end up that the time for testing and detailed evaluation will not suffice. To manage
this risk, a clear project timeline, which will outline the activities that will be performed,
timelines for data preparation, model development and testing will be developed.
Communication with the end supervisor will be on a routine basis to check on progress made
and to address problems as they crop up. Further, there will be contingency plans with
possibilities of delays in accomplishing the activities.

4
 3) Technical Risks:
The technical risks are related to the quality of data input; the performance of the convolutional
neural network (CNN); and the capacity of the hardware devices. If the data set being used is
either incomplete or of a very low quality then this is likely to affect the performance the model.
To deal with this, an initial data cleaning process and then data augmentation techniques will
be used in the experiment, and the results obtained will further be validated with other datasets
and analyses. Another risk is that deep learning to train models depend on the availability of
GPU. Nonetheless, in the event that available hardware resources are sparse, time spent in
training might be greatly affected. To this end, cloud computing systems including Google
Collab or AWS will be employed to meet the required computational resources requirements.

References in IEEE format (Approx 5 References)

1. A. Krizhevsky, I. Sutskever, and G. E. Hinton, "ImageNet classification with deep convolutional

neural networks," in *Proc. of the 25th Int. Conf. on Neural Information Processing Systems
(NIPS)*, Lake Tahoe, NV, USA, Dec. 2012, pp. 1097-1105.

2. D. P. Kingma and J. Ba, "Adam: A method for stochastic optimization," in *Proc. of the 3rd Int.
Conf. on Learning Representations (ICLR)*, San Diego, CA, USA, May 2015. [Online]. Available:
https://arxiv.org/abs/1412.6980.

3. L. S. B. Wu, "Brain tumor detection and classification using convolutional neural networks,"
*Journal of Medical Imaging and Health Informatics*, vol. 9, no. 3, pp. 511-517, Mar. 2019. doi:
10.1166/jmihi.2019.2653.

4. Y. LeCun, Y. Bengio, and G. Hinton, "Deep learning," *Nature*, vol. 521, no. 7553, pp. 436-444,
May 2015. doi: 10.1038/nature14539.

5. M. D. Zeiler and R. Fergus, "Visualizing and understanding convolutional networks," in *Proc. of

the European Conf. on Computer Vision (ECCV)*, Zurich, Switzerland, Sep. 2014, pp. 818-833.

These references cover fundamental concepts of CNN, optimization algorithms, and their
application in medical image processing, fitting well with the techniques used in this brain tumor
detection project.