Scribd
Upload
26 views3 pages

Brain Tumor Segmentation

The document describes developing a CNN model using the U-Net architecture to detect brain tumors in MRI images. The model achieved 90.89% accuracy and 48.62% loss on a dataset of 484 patients. Preprocessing steps like normalization and augmentation were applied to the data. The model exhibits potential for clinical use but could be improved by addressing class imbalance and architecture tuning.

Uploaded by

Dwaipayan Nandi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
26 views3 pages

Brain Tumor Segmentation

The document describes developing a CNN model using the U-Net architecture to detect brain tumors in MRI images. The model achieved 90.89% accuracy and 48.62% loss on a dataset of 484 patients. Preprocessing steps like normalization and augmentation were applied to the data. The model exhibits potential for clinical use but could be improved by addressing class imbalance and architecture tuning.

Uploaded by

Dwaipayan Nandi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Brain Tumor Segmentation using CNN

Introduction
In the realm of medical imaging, accurate and efficient brain tumor detection plays a pivotal
role in timely diagnosis and treatment. This project aimed to develop a Convolutional Neural
Network (CNN) model for brain tumor detection using the UNET architecture. The dataset
comprised 484 patients' data in NIfTI format, with image sizes of (240, 240,155). The model's
accuracy was 90.89%, with a loss of 48.62%.

Methodology

i) Dataset Pre-Processing
The foundation of any successful machine learning project is the quality of the
dataset. In this project, the dataset consisted of 484 patients' brain images in NIfTI
format, each with dimensions (240, 240, 155). Prior to feeding the data into the CNN
model, a series of preprocessing steps were carried out. This included data
normalization to bring pixel values within a certain range, and conversion of NIfTI
files to numpy arrays. Furthermore, to prevent overfitting and enhance model
generalization, data augmentation techniques were applied. Random transformations
such as rotations, flips, and zooms were employed to create additional training
samples.

ii) U-Net Model Architecture

The U-Net architecture, known for its success in medical image segmentation tasks,
was adopted for this brain tumor detection project. The U-Net architecture consists
of an encoding path that captures relevant features from the input image and a
decoding path that reconstructs the segmented output. Skip connections between
corresponding layers in the encoding and decoding paths allow the model to retain
fine-grained details during upsampling.

iii) Training
The training process involved feeding the preprocessed and augmented data into
the U-Net model. A suitable loss function, such as binary cross-entropy, was
chosen to measure the discrepancy between the predicted segmentation and the
ground truth. The optimization algorithm used for updating the model's parameters
was Adam. Adam's adaptive learning rate and momentum calculations help in
efficient convergence and handling of various types of data.
iv) Evaluation
The evaluation of the model's performance was carried out using a separate
validation dataset. During training, at the end of each epoch, the model's
performance was assessed using metrics such as accuracy, precision, recall, and F1-
score. These metrics provided insights into the model's ability to correctly classify
brain tumor regions while avoiding false positives and false negatives. Additionally,
the loss value indicated how well the model was converging towards the optimal
solution.

Results
The developed CNN model exhibited promising results. With an accuracy of 90.89%, it
showcased a significant capability for accurately identifying brain tumors in the given
dataset. The loss of 48.62% indicates that the model was able to effectively optimize its
parameters to fit the training data.
Discussion
The achieved accuracy underscores the potential of CNNs in medical image analysis,
particularly in brain tumor detection. However, the relatively high loss suggests the presence
of challenges in the dataset, such as class imbalance or the complexity of distinguishing tumors
from healthy brain tissues.

Future Work
Future endeavors could focus on addressing the limitations observed in this project.
Exploring advanced data augmentation techniques, investigating class imbalance mitigation
strategies, and fine-tuning the model's architecture could further enhance its performance.

Conclusion
This project demonstrates the potential of CNNs, specifically the UNET architecture, in
accurately detecting brain tumors from medical images. With a notable accuracy of 90.89%,
the model exhibits promise for clinical application, potentially aiding medical professionals
in timely diagnoses.

Acknowledgement
We extend our heartfelt gratitude to Dr. Debashis Nandi and Mr. Rohit Agarwal for their
invaluable guidance and support throughout this project. Their expertise and insights were
instrumental in the successful completion of this endeavour.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy